Rapid7 Vulnerability & Exploit Database

Sun Patch: VxVM 3.2s9: general patch for Solaris 9

Back to Search

Sun Patch: VxVM 3.2s9: general patch for Solaris 9

Severity
4
CVSS
(AV:L/AC:M/Au:N/C:P/I:P/A:P)
Published
02/26/2004
Created
07/25/2018
Added
05/03/2006
Modified
01/30/2015

Description

From Sun Patch 112385-06

Sun has released a security patch addressing the following issues:

NOTE: VERITAS Incident Numbers are included in parenthesis.
4479879 (116784) vxdiskadm option 20 (yes) will change disk status from 
		 "error" to "offline"
4490646 (114693) Various issues with vxdmpadm usage messages and man page
4773633 (96801)  CVM: vxconfigd: WARNING: library and vxconfigd disagree on 
                 existence of client
4820436 (128610) VM/licensing functionality is not fully enabled in leadville 
		 configurations
4846600 (129377) NODE_EXPLICIT and DMPNODE_EXPLICIT not get reset after migration
4869501 (108322) (131368) CVM: VOLOBJ_MV_FIRSTW flag incorrectly remains set 
		 causing close to hang
4875401 (126492) vxedit rename can degrade performance of VM
4881620 (131823) VxVM: VM is recognising disks from other ( V880, 3510 etc) enclosure
		 as SENA disks
4882902 (128502) vxconfigd coredumps when photon enclosure is replaced due 
		 to NULL pointer
4884319 (136641) S85vxvm-startup2 will copy new shared libraries while 
		 vxconfigd is running
4886361 (131823) VxVM: VM is recognising disks from other ( V880, 3510 etc) enclosure
		 as SENA disks
4887356 (138553) CVM/SC30: Moving of quiesce ioctl to step1 to modify 
		 the join behaviour
4892981 (126223) vxassist mirror fails to mirror vlumes of size >1TB
4900508 (133579) Update VxVM JBOD info to include non-Seagate JBOD FC disks
4903611 (134055) Unwanted console messages for plexes/volumes on CVM Master node
4908473 (128878) CVM: Oracle Resilvering: Need to not duplicate effort in setting 
                 MR mode
4911562 (100379) VxVM: Diskgroup creation is failing if two vxdg commands are given 
                 simultaniously
4928807 (136853) After installation of patch - some libraries cannot be linked
4930738 (118641) vxconfigd coredumps when running "vxdg -g <group> -k rmdisk 
                 diskname"
4937259 (128878) CVM: Oracle Resilvering: Need to not duplicate effort in setting 
                 MR mode
 
(24329)  vxconfig help message incorrect
(34523)  Need to increase VOLD_MAX_REQUEST to handle larger szie request to vxconfigd
(81289)  vxrecover does not output appropriate errs and always has 0 return code
(101657) Replacing a disk can violate a photon array license
(105230) SOLARIS 9: pkgrm VRTSvxvm panics the system
(106277) vxassist maxsize|make takes more time to complete (15Hr sometimes)
(113157) vxconfigd core dump due to stack corruption caused in dg_creat_internal()
(118770) panic in scsi module during pkgadd
(123003) volume stuck in SYNC with no sync action
(123102) Race condition may lead to 'vxdmpadm disable ctlr' hangs due to missing 
         signal from 'gendmpiodone()'
(123806) FMR2: multiple snapshots of a volume taken in the same command are not 
         tracked correctly
(123974) T3 support: DMP need to use FNDELAY to allow sense data to be captured
(125951) VM re-encapsulate root fails silently and system unbootable if vtoc already 
         has private slices
(127515) CVM fmr/fmrlogsz.tc fails occasionally with master message
(128216) vxrecover core dumps if given non-existent object as input.
(128378) GLM slave node stuck in vxg_api_deinitlock via volcvm_gab_joindone
(128434) rem_drv exit status may not be reliable
(128571) potential deadlock in vxdmpadm enable path
(128983) vxdisksetup is giving "UNKNOWN ERROR WHILE WRITTING VTOC" when publen and 
         privlen specified
(129080) memory not unmapped after a mmap in crc_file()
(129174) According to Sun, VxVM 3.2p4 can not depend on Solaris patch 108827-XX
(129295) Sol: encapsulation does not check for rootvol in the group
(129462) FMR-CVM - Map manipulations are not properly interlocked with I/Os on slave
(129629) VxVM is not starting with VE_LOG_BAD error on second reboot after 3rd 
         mirror break off from rootvol.
(130324) Unwanted dmp debug messages are periodically appearing on console
(130774) FUJITSU claim_device: fjget_path_mode failure
(131595) Cannot set spare flag via vxdiskadm.  dmname "d1" matches daname of 
         c#t#d1ds2"
(133070) The Fujitsu-Siemens Primepower arrays no longer have the eeprom set properly
         when encapsulating
(133481) vxvol core dumps in common_start_resync
(134236) vxassist relayout not handling VE_RESTART at all places
(134510) CVM: neighbor field not set properly after a reconfiguration in certain 
         conditions
(134946) Sol: SENA is not recognised when MPxIO is enabled on the machine
(135454) vxsync: /etc/mnttab: Too many open files
(135682) Data corruption with DRL if system crashes before klogging plex detach
(136128) Hitting dmp_io_failed_threshold should log a console message
(136224) dmp_io_failed_threshold default of 600 seconds is dangerous
(136404) No primary path failback happens with A/PG arrays.
(137176) TED assert when doing an IO from a slave in vol_open_device
(137452) Snapshot of multiple volumes with ufs and vxfs atop yields error msgs
(137641) vxconfigd core dumps in dmp_get_exclude_info() due to NULL pointer
         reference
(137670) TED_ASSERT in vol_open_device and vol_close_device are not needed
(138205) CVM: slave join fails due to unexpected change in errno value
(138285) IBM SANVC luns with VPATH behaving unpredictably on Solaris VM3.5 and VM 3.2
(138389) performance degradation while running a volume sync
(138740) copystr() should handle errno == EAGAIN and ENOMEM from xcopy().
(139796) vxcheckasl giving Error opening shared library libvxsena.so
 
 
(from 112385-05)
 
4233507 (73476)  vxvm should periodically dump disk group configurations to a file 
4421145 (61755)  typo in vxdmpadm help output for option getsubpaths
4487591 (76135)  vxddladm man page should describe ddl better 
4479302 (95151)  modified libvxvm_get_disks() in libvxvmsc.so
4486602 (118007) vxinstall does'nt display "excluded disks" message for disks
         	 in enclosures
4490375 (82309)  vxddladm excludearray vid/pid shouldn't indicate entire
                 library has been excluded
4490646 (114693) Various issues with vxdmpadm usage messages and man page
4491048 (114033) man page for vxbootsetup refers to 'disksetup' and 'edvtoc' 
		 instead of 'vxdisksetup' and 'vxedvtoc'
4491997 (76135)  vxddladm man page should describe ddl better
4501921 (113692) vxunroot does not work as documented
4506093 (81823)  CVM: vxdg -k rmdisk fails
4527907 (115102) request to backport fix from 94043 to 3.2 and 3.1.1
4628999 (94225)  Excluded pathnames were displayed contiguously in a single line
4629246 (114060) man page for vxplex is incorrect for the 'cp' section ; volume
         	 should be enabled not disabled as documented
4629303 (113193) Cannot change log types from DRLSEQ back to region (DRL)
4630477 (101371) bogus device in `vxdisk list` output after replacing a disk drive
4643108 (117826) Document that 3.2 patches are common for all localizations
4648109 (43033)  SOL: dmp_per_cpu_stats_t should be 128 bytes long
4655082 (98658)  vxmirror using Sun StorEdge S1 drives create incorrect root
                 devaliases (vx-rootdisk and vx-disk01)
4656592 (108677) CVM volumes disabled after scshutdown and boot up on 
		 all cluster nodes
4671086 (110341) CVM: SUNW wants vxclust to not start vxconfigd with logging
4685297 (100354) world writebale files in /var/vx/ and /etc/vx - permissions
		 need to be modified
4711196 (114288) Encapsulation creates a public region which overlaps private
                 region
4712148 (121955) vxdiskadm #6- cannot mirror rootdisk to an enclosure based /name
                 target disk with 3.2patches
4712376 (118011) Strange notification mesg from vxconfigd when enabling MPxIO
		 ( atf vendor_info: readlink fails )
4749183 (112049) With just Photon license  (98) - RAID5 log on a non-photon disk
		 is allowed
4753143 (109148) VxVM does not generate photon license with MPxIO enabled
4757423 (115645) VxVM vxassist -H indicates "Incorrect usage"
4767258 (112554) vs_pl_cnt (vol_summary's field for plex count) gets negative 
		 values on SLAVE when using vxmake to create sd/plex/vol and 
		 then deleting the volume
4769704 (114479) Request for relevant error messages: specifically when
		 vxdiskadm option 5 is used on duplicate disk ids
4770528 (111945) vxdiskadm option 5 - need to explicitly specify device_name 
		 usage ;erroneous error message and output format
4775892 (102494) '/usr/sbin/prtconf: getexecname() failed' on system bootup
4780047 (112593) Incorrect devalias boot path on Jasper HBA
4780146 (112279) Problems with enclosure based naming after replacing 
		 physical disks
4784298 (93770)  vxdiskadm (disk.mirror) cannot mirror disks with enclosure
		 based naming
4788268 (124695) CVM: disk group autoimport flag is turned off after 
		 Master takeover
4792191 (70508)  'vxdisk list' should display the disk to enclosure relationship
4793932 (117922) Add to VM 3.2_p4 README info about 3.2_p3 disabling VM 
		 because Photon license is no logner free on a SunFire V880.
4794148 (28606)  VxVM does not correctly set dump device
4800160 (115323) vxddladm during VxVM 3.5 vxinstall takes an unusual amount 
		 of time on hosts with maserati devices 
4817095 (118065) Loss of storage connectivity to CVM master causes all nodes 
		 to get I/O errors 
4819716 (118003) man page for vxdisksetup needs to be updated with information 
		 about 'old_layout' option
4822685 (118015) "vxdisk list" does not report multiple failed/removed disks
		 correctly
 
(24412)  subvolume SIO's not being freed
(55730)  rootability error msg with vxunroot
(66749)  pkgchk VRTSvxvm error; vxserial not properly linked to vxlicense
(71594)  VM - hitachi 5800 (DF400-F) os warning msgs with vxdctl enable
(72972)  VVR: vxprint -lP does not print the correct protocol for older DGs < 90
(73964)  VVR  vxassist relayout of an srl should not be allowed
(84648)  mread ioctl is making kernel to panic 
(88790)  DMP doesnt migrate disabled other_disk dmpnodes properly after they 
    	 become available
(90840)  vxrelayout dumps core with Segmentation Fault when striping column
(91319)  vxinstall not allowing an enclosure name to be renamed
(93840)  the requirement of stopping and starting the whole layered volume
(94409)  VVR  Secondary panic during tcl's
(95032)  SOL: vxdmpdebug failing with "adb: failed to read from address 0:
	 no mapping for address"
(95665)  vxdisksetup needs a slice 2, but doesn't detect if its not there 
(96437)  Occasional 'vxddladm addjbod....' core dump
(97089)  dmp macros not dumping the last dmp node in lvl1_tab
(97192)  ddl_add_support() ought check return value of ddl_vendor_info()
(97270)  VVR: vxrlink stats shows negative stream errors
(97656)  VVR : TCP : RT(usec) should be changed to RT(msec) in rlink stats
	 output
(99600)  vxdiskadm - adddisk option failed to add disk into the shared DG 
(101025) Mirroring fails on disk when enclosure name is in used
(102668) VM - array.info file is set @ 555 permissions
(102738) Make vxassist use SAL XML spec 1.2 to work with SPC 2.1.1 onwards
(102913) subcmds "mirror" and "move" and others in vxassist hit 
	 VE_NOTRANS sometimes
(103649) Volume badlog flag set to 'on' , with valid log subdisk available
(104742) vxasldebug uses absolute pathnames to tar up /etc/vx/lib/discovery.d/* 
(105234) drl: vxassist -o ordered make requires logdisk arg too, contrary to 
	 sag doc 
(105598) vxdisk list fails if disk media name is similar to a non-existant 
	 CTD device
(105659) FMR2 - Do not allow the resize of DCO volume
(105946) VVR: rlink disconnects while doing autosync
(106277) vxassist maxsize|make takes more time to complete (15Hr sometimes)
(107746) VVR: vxrvg recover takes long time since SRL header is not flushed
(108389) i/o hang on volumes with drl
(108422) vm is trying to create volume even if fmr licence is not there
(108441) vxprint taking a very long time to complete
(108867) vxconfigd dumps core during 'vxdctl enable' after pulling out a cable 
	 from a SENA AA array 
(109231) plexes with SNAPDONE state are not detached on i/o error
(109656) Panic in hitachi_failback_lungroup for HDS A/P 9200 Array with 
	 VxVM
(109778) CVM support for Hitachi 9500 DF600 APG
(109851) vxassist convert applying wrong trigger pt and not checking for 
	 simple volume
(110265) VVR: associating a volume after role-reversal leads to config error
(110319) VVR: Remnant config info on secondary after primary rvg is deleted can
	 cause data corruption
(110508) VVR: Panic in nmcom_server_start due to NULL msgb returned by
	 kmsg_sys_rcvudata
(110536) VVR: Data loss in autosync of vol w/regsz > 512 blks & rlink
	 discon/connects
(110742) Having problem adding disks to an existing shared diskgroup 
	 using vxdiskadd
(110851) panic in dmp_migrate_node() routine
(111119) Multipathed rootdisk will fail to import  with 'duplicate diskid'
         message, in VM 3.2_p2 and above.
(111143) VVR : changing packet_size during replication leads to data corruption
	 and/or panic
(111288) postinstall script does not update library file list for pkgchk
(111894) ROUND read policy not used anymore when a plex is detached 
	 (3-plex volumes)
(112517) vxsync errors needs to be changed to a warning
(112686) VVR : replication hangs if packet_size > 65504 is set and an io of
	 size >= 64K is given
(112770) SCSI version for SENA disks does not come correctly
(113279) pkgrm -R should not unload (local) vxvm modules 
(113838) vxassist getting error message when using the '-o ordered' option
(114633) vxrelocd fails after switching to enclosure based naming 
(115325) VVR: vxrlink att fails displaying negative number for volumes over 1TB
(115520) vxconfigd dumps core in dmp_get_events () for DMP_DA_ADD_EVENT if the
	 enclr is not present in the attr_list
(115607) VVR: Panic in vol_dcm_set_region during DR testing
(115771) vxdg deport hangup after dg disable 
(116105) panic while creating a snapshot of raid5 volume
(117013) Inappropriate error message with vxdisksetup on an invalid device
(117388) Panic and filessytem corruption after doing "rm *" during a raid5 sd
	 move operation
(117568) use this incident to detect and fix mem leaks in vxconfigd
(118064) FMR: ted assert _vol_syncwait:1a in voldco_reset_state
(118172) relayout fails with VE_DG_VERSION even though dg version is at 80 
(118186) vxtrace help does not explain all options of the command
(118243) Change LUN ownership support for Fujitsu GR710, GR720, 
	 and GR730 arrays 
(118609) vxbootsetup prints error message test: argument expected 
(118770) panic in scsi module during pkgadd
(119565) DMPNODE not migrated when array type changes 
(119864) panic in dmp_dev_t_to_nodep()
(119998) relayout (in 3.2p3) can exhaust dg config space
(120377) memory leak in (vxconfigd) priv_scan () during diskgroup import
(120459) update modinfo and pstamp for VM 3.2t Patch 4
(120808) VxVM 3.2p3 vxvm-startup2/vxreattach performance degradation on boot
	 for large configurations
(120830) Generic support of A/P arrays in CVM/DMP
(121428) pkgrm runs w/minor complain when removing VM 3.1.1 for an upgrade
(121957) vxconfigd core dumps while doing bascii_copy() in a5k_get_da_sno with
	 SunAP 2.3.1
(121962) vxconfigd is exiting with symbol strfind:referenced symbol not found
	 error.
(122528) vxdctl enable hangs in ddladm3.tc hangs on latest 3.2
(122982) patchadd fails because of a typo
(122984) preinstall fails as required SUN patch is superseded
(123311) VM: vxconfigd -m disable gives core
(123974) DMP need to use FNDELAY to allow sense data to be captured
(124315) inst.top.sh refers incorrectly to JBOD and OTHER
(124991) vxconfigd dumps core during zrelayout/crash_dbg.tc
 
 
(from 112385-04)
 
Readme changes and rebuild.
 
 
(from 112385-03)
 
4295380 (28606) VxVM does not correctly set dump device
(63819) ./scripts/admin/volprint/t_flag.tc#1 FAILS on debug only
(66490) ddl should remove the second path info from vxdisk list after the array
        has been claimed under JBOD0 by the vxddladm addjob operation
(74919) r5log.tc#8 failed
(75389) Unable to remove a disk containing only a subdisk and/or plex
(76502) VVR: vxibc_unfreeze.tc fails and rlinks go connect/disconnect.
4148664 (76636) root disk encapsulation should fail when bad license key used
4874143 (78526) vxevac not handling subdisk states manipulated by user correctly
(80615) mirrored volume is not disabled when it lost the last ACTIVE plex during 
        resync
(81182) dmp restore daemon's 'skip controller decision' in check disable policy
(85883) vxencap.sh fails on Solaris 2.6 after fstyp to non-existent slice
(87807) I/Os stuck in VxVM I/O mempool waitq
4633949 (87863) VxVM 3.2 gives free PHOTON license on V880 system
(87931) RECORD fragmentation may lead to missing of record becuase of the block
(87968) FMR: Data corruption if m/c crashes after DETACH klog, and before accumulator
        update
(87975) vxdmpinq - add support for detail display of scsi inquiry data
(89014) vxunroot does not work when /usr/local on disk
(89095) vxconfigd out of memory; vx commands get "Memory allocation failure"
(89170) Policy to handle duplicate diskids
(90069) DDL: ASL packaging changes
(90710) replacing a disk when we have multiple DRL's does not clear FAIL state
(90828) ../switchout/raid5/volsd/mv/io_fail_dbg.tc#2,#5 fail on ga3.5_maint
(91698) vxdisksetup -i uses -f option of vxdisk
(91793) CVM callout table flooded on single CPU machines
(92157) Luns>10 assigned to wrong controller on the HDS 9200 array
(92232) kernel/ioctls/lv_oradrl.t fails when vol_maxio tunable is increased
(92350) coverage/coverage_F.tc#2 fails when vol_maxspecialio tunable increased
(92524) CVM: DCO update protocol does not handle cluster reconfiguration properly
(92581) vxunroot does not convert all volumes to partitions
(92817) incorrect argument passed to sema_p
4504662 (93080) vxdisk list reports disk status incorrectly
(94911) vxvm is excluding a disk that not specified in vxvm.exclude
(94958) VVR  ./stress/autosync_largevol.tc#13 hung primary node
(95241) vxeeprom creates incorrect boot-device string for a5k
(95545) CVM two nodes of a 8-node cluster never join - master vxconfigd gets 
        'kernel_fail_join' error and hangs
(96137) "vxdg -k" should be prevented on encapsulated rootdisk, otherwise system 
        will be unbootable
(96364) VVR: vxrlink pause hangs for Secondary rlink
(97339) 32p2.5  - mirroring fails w/errors on a smaller target disk
(97508) DISK-SUBSYSTEM entries on XP array shows up in "vxdisk list"
(97600) ATF issue with boot volume on CLARiiON FC4700
(97947) VVR: vol_rp_ack_timeout should check the first sio for timeout.
(98045) vxassist relayout does not preserve usetype gen
(98221) VVR: Replication hung while running autosync_rlinks.tc
(98366) VVR: dcms on stripe-mirror volumes not working properly
(98486) "vxassist make" fails to create volume with VE_NOTRANS
(98880) vxassist make of mirrored multi-col volumes should initialize cols in 
        parallel
(98891) VVR : memory leak
(99015) vxunroot doesn't remove those commented entries in /etc/vfstab
(99087) VVR: Replication hangs because rp_config_status.done field gets cleared 
        during transaction
(99216) Initiator hang when disable the primary controller
(99611) stress/vmconfig/vmconfig05_dbg.tc fails to cleanup properly
4666269 (99652) Why Volume Manager needs to forceload SUNW,isptwo?
4693346 (99652) vxvm 3.2 tries to forceload non-needed drivers on 2nd reboot of boot 
                encapsulat.
(100170) After vxinstall, swap device is not being set, on boot getting 
         "cannot open /etc/vx/reconfig.d/saveconf.d/swapdev".
(100375) VVR: Replicas differ after Primary failover during resync
(100581) Add CVM support for Hitachi A/P arrays
(100852) VVR: rlink disconnect due to ack timeout on update message
(100866) Sol+cfs+stress+reconfig fails to make progress due to thread stuck in 
         vol_kmsg_send_wait.
(101177) vxreattach recovers disks with duplicate ids
(101343) 3.2 vxinstall could cause mismatch between publen and partition table
(101637) memory corruption in vol_dump_disks
(101825) CVM SC3.0 vxclust step2 poll timeout prevents third node from joining
(102642) VVR  hdr/dcm-header does not flush after srl overflow
(102777) vxdisksetup fails if disk access name matches an existing disk media name
(103005) vxassist mirror sometimes fails silently
(103454) In vxdiskadm (option 5) cannot replace a failed pre-3.2 disk with new disk 
         layout (CDS), because of smaller publen.
(103807) vxassist dumps core.
(103917) "vxassist make" fails with VEX_TIMEOUT
(104091) VM: Quick Install in vxinstall does not exit properly if no disk were 
         selected.
(104177) ASL: ASL packages are modifying /etc/vx/aslkey.d and  
         /etc/vx/lib/discovery.d directories from 755 to 775,
(104246) DMP issues with VM 3.1.1 and SAN appliance
(104384) diskgroup import is slow for configuration with large number of disks
(104438) vxplex fails to use FMR to resync offline plex
(104444) mirroring of root disk failed
(104541) CVM/FMR vxsync never worked in a cluster
(104656) Group failover failure due to determining incorrect num_luns on a controller
(104673) ASL: vxasldebug command fails with "cannot create" error message, DATE is 
         incorrect.
(104815) Add Solaris 9 support to ga3.2_maint
(104916) VxVM on Solaris: Timeouts using lbolt can be too short
(104926) vxassist make logtype=drlseq silently fails to make DRL
(105116) VXVM 3.x hotspare of rootdisk fails if disk failure is after a controller 
         failure
4724737 (105185) libvxvmsc.h header file missing from VRTSvxvm package
(105375) CVR for rusio iocount is incremented but never decremented
(105447) libvxvmsc library file is missing in 3.2p2. Need to make sure it is in 3.2p3
(106045) increase VOLDIPC_MAX_FD to avoid VxVM commands hang
(106379) vxdg adddisk allows duplicate diskid in DG
(106396) System panics during SPFS array cable pull testing
(106401) update modinfo and pstamp for VM 3.2t Patch 3
(103917) "vxassist make" fails with VEX_TIMEOUT
(104916) VxVM on Solaris: Timeouts using lbolt can be too short
(106379) vxdg adddisk allows duplicate diskid in DG
(106401) update modinfo and pstamp for VM 3.2t Patch 3
(106695) Error message due to RDAC device
(106879) vxunroot fails to see any other vols than / swap usr opt var home
(106907) VxVM doesn't sync the whole plex after replacing disk
(107261) I/O stops after a drive was pulled on RAID5 volume
(107284) Possible wrong I/O count on objects because of not handling 
	 dg_objcpulist correctly
(107382) FMR2: a system crash after a disk is replaced by some other disk may 
	 result in data corruption.
(107473) SNAPDONE plexes are not resync'd after a crash
(107656) vxinstall should remove install-db if just using old configuration
(107841) update modinfo for pointpatch
(108091) Panic during encapsulation of an AP controlled bootdisk in VM3.2
(108138) FMR doesn't resync properly after disk replacement
(108162) vxassist dumps core during relayout
(103206) Data coruption in FMR (102132)
(103209) update modinfo and pstamp for point patch for 3.2t_p2.5 to contain
         incident 103206
(29953) vxassist help space shows incorrect numbers
(30225) No need to call vol_mv_stable_reserve() if not stabilizing
(31590) EMC:libarray.so stub library has inconsistent return codes
(32999) EMC: vxinstall fails while adding all disks
(35074) EMC: vxdg deport/destroy allows you to deport DG with BCV attachments
(43263) General updates and corrections to VM man pages
(49711) SRVM: Application hangs when doing writes on a detached rlink.
(50206) EMC: vxassist move fails for volumes on non-EMC disks with verify option
(56510) the current certified version of ATF 1.3.1 with DMP is too old
(58021) EMC: Incorrect use of vxsym_errno in vxvm
(60034) EMC: Support new options in vxdiskadm, vxdiskadd, vxresize, vxrootmir and 
        vxmirror
(60732) Check-in DDL and independent device naming man pages for 3.2 release
(61829) root disk encapsulation program corrupt the partiontion table.
(70415) DMP cannot hardcode pub/priv partitions
(70611) fix vxencap issues found during testing for normal disk
(71336) VVR: vols go into SYNC state during recovery if rvg is in PASSTHRU mode
(71362) vxnotify does not send event notification at time dmp path os disconnected
(71613) VVR TC: vvrtc failure on the latest build
(72734) VM 3.2p - The man page for vxddladm and the vxddladm help text do not match 
        well
(72853) VVR  panic on secondary during mrs_recover.tc#1
(73510) problems with dcm and layered volumes
(73836) VVR: rlink enabled even if it was detached when there's srlheader error
(74085) Add DDL support for Fujitsu arrays
(76063) Upgrade process does not comment out VxVM volumes in /etc/vfstab
(76498) USETYPE of var,usr,opt changed to gen in vxvm 3.1.1p2 after using vxinstall 
        to encapsulate the boot disk
(76940) VX_GET_IOPARAMETERS ioctl returns invalid dev_t on cvm slaves
(77239) importing a diskgroup with a duplicate dgid doesn't generate error
(77259) VVR: ted assert when running aixvm/tot_merge on solaris.
(77987) No more lite license without a key
(77993) VVR: SRL failure may result in inconsistent data
(78595) EMC: P3A: device group should be renamed whenever a diskgroup gets renamed
(79563) switching to enclosure based DA names creates duplicate records.
(79655) CVM: ora.tc consistently hangs on 4 node GAB cluster
(79735) FMR2:Crash recovery for snapstart
(80587) VVR: vxrvg -M makeprimary doesn't work properly
(80735) VVR : panic on solaris while running srvmtest on tot
(81083) FMR:panic from snapprint in 3.2
(81763) EMC: Update the Testplan for handrun TCs and correct some TC scripts
(81867) vxstat -i should flush stdout
(82369) FMR:vxassist fails silently to remove a dco log volume
(82974) VVR: Too many memory errors on draining SRL when using TCP
4506689 (83191) vxdiskadm replaces disk but errors with UNKNOWN ERROR WHILE WRITING 
                VTOC
(83351) libfdisk is out
(83386) vxprint -e takes long time to execute
(83557) Panic in do_xor() because of alignment error
(83719) change SIG dependency name in VxVM spec file
(83724) Support for new licensing utility package.
(83928) FMR2: Proper cleanup to be done while aborting a snapstart and snapback
(84371) P3A: the DGSJ of the Toolkit operated disk groups need device-group 
        reconfiguration as well
(84375) vxassist snapprint doesn't work when a disk group is split
(85139) Data corruption problem in FMR
(85288) changes reqd for code that creates /dev/vx/rdmp /dev/vx/dmp links
4533901 (85333) VxVM and AP231 co-existence fails.
(85460) VVR : TCP : should free conn_req in case not able to alloc conn_ret
(85560) vxdisksetup gives confusing error message
(85908) VVR : nio_usec_tstamp can overflow and become -ve.
(86160) vxdmpadm -f disable path=c#t#d#s2 prints incorrect error about NON-rootdisk, 
        saying it IS a rootdisk
(86164) If an operation in vxdiskadm fails and nothing is changed we should not run 
        'vxdctl enable' when exiting.
(86291) VVR: Replicating betn. diff VVR vers results in connect/disconnect
(86340) added disks are not updated
(86594) Wrong stats maintenance in voldisks
(87403) VVR: panic in nmcom_get_checksum if packet_size of rlink is not 4 byte 
        aligned.
(87461) removal of a device from a STD dg which is attached to a BCV dg was 
        successful with enclosure based naming scheme
4616042 (87645) VxVM is not detecting all slices created on T3B with T3B_FW2.1
(87821) DDL: Built-in ASLs of VxVM should build under VRTSasld also
(88585) vxdmpadm getdmpnode nodename
(88847) vxconfigd panic with alignment error, when dmp_debug_level set to 9, 
        gen_aa_select_path
4505811 (89089) 3.1.1/3.2: /etc/rcS.d/S85vxvm-startup2 should copy additional 
                libraries
4545695 (89089) VERITAS Volume Manager commands may break after upgrading to 
                Solaris 8 Update 7 or installing Patch 108827-15 or Later
4627914 (89089) vxvm-startup2 vxvm 3.1.1 does not copy all vxvm required libraries
4634822 (89089) VXVM-patch should copy needed libraries upon installation
(89399) Arithmetic exception with relayout on large stripe width plex
(89428) VVR: srl is inconsistent but rlink is not detached
(90451) ADB macros used by VRTSexplorer need updation to collect complete dmpdb
(70904) vxedit -v set owner=xxxxx doesn't work in Volum Manager 3.1.1
(80961) vxdmpadm listctlr all and getsubpaths dmpnodename lists the CTLR-NAME 
        incorrectly after enabling MPXIO
4615526 (83140) voladm_get_disk_dmname routine uses bogus default disk
(85892) VVR io stops at 99% full srl in asyn with dcm
(85994) VVR: Disallow associating an open volume to a secondary rvg
(85998) VVR: Disallow attaching a secondary rlink if any associated data vol is open
(88498) FMR - performance improvement in DCO updates
(88545) VVR: Disallow associating an open volume as SRL
(88794) io error on a volume while doing 'vxdisk scandisks' after 'vxddladm 
        includearray libname=xxx'
(89170) Policy to handle duplicate diskids
(90335) dd hangs when cable is pulled from hitachi 9200 array
(90471) ddl and dmp APIs should close files/devices opened before return
(90533) CVM volume stuck in NEEDSYNC for SC3.0
4545695 (90583) vxconfigd core possiblity: memcpy may read past the end of the 
                source string
(90728) ATF: VxVM does not see the failed over ATF LUNs after rebbot
(90956) update modinfo and pstamp for VM 3.2 Patch 2
(91452) SEAGATE/HITACHI devices should be added as JBOD in postinstall scripts for 
        Solaris
(92725) VVR  vol_rp_delete_port panic during mrs_recover.tc#1
(92921) VVR  Primary node hangs during 
        ./scripts/offhost/offhost3.tc#2:offhost-backup-2:
(93035) VVR: KMSG/TCP: Add support for option management -- kmsg_sys_optmgmt
(93127) VVR: TCP: Increase snd and recv buffer space to improve performance
(93493) always add SUN Vicom arrays as jbod
(93495) update modinfo and pstamp for point patch for 3.2p1 to include 90533
(43263) General updates and corrections to VM man pages
(56510) the current certified version of ATF 1.3.1 with DMP is too old
(70611) fix vxencap issues found during testing for normal disk
4419634 (77622) VxVM - concat-pro & stripe-pro volumes do not align subdisks on 
                cylinders.
(78995) vxencap man page doesn't describe [-s priv_reg_len]
(79028) vxbootsetup man page should refer to troubleshooting guide for the 
        'Recovery from Boot Disk Failure' chapter
(79477) remove bogus messages while building VRTSvxvm rpm
(79845) update modinfo and pstamp for VM 3.2t Patch 1
(81395) Man page changes requested by Japanese l10n
(82247) 3.2p1 - cannot remove/replace failed rootdisk via vxdiskadm
(82965) pids for daemons have changed in 2.4.9
(83010) New version of rpm splits rpm and rpmbuild
(83257) Under certain configuration, vxconfigd can core dump when doing enclosure 
        renaming
(83351) libfdisk is out
(83698) 3.2tp1.5 causes hds5700 to migrate from HITACHI to other_disk category.
(83711) vxvm-reconfig tries to cat non-existing files
(58257) vxio and dmp drivers must fail insane disks without retrying i/o to them
(72742) VM 3.2p beta: unable to mirror rootdisk via vxdiskadm
(79845) update modinfo and pstamp for VM 3.2t Patch 1
(82076) Addition of new path is failing in 9200 array
4469908 (70451) vxprint outputs gives impression that volumes are not correctly 
                created
(71548) Add Autoraid suport to DDL
(70451) VM 3.2beta: vxprint output of a stripe-pro vol erroneously lists plexes with 
        no associated sub-disks
(71548) Add Autoraid suport to DDL
(72686) vxtask list shows wrong object names for Solaris 64 bit kernel.
(73679) Hitachi 9200 command device not excluded during discovery
4505737 (79192) VxVm reading from STALE plexes / Possible Cause of Silent Data 
                Corruption on VxVm mirrored volumes
(79845) update modinfo and pstamp for VM 3.2t Patch 1
(79959) I18N checkin of help files, and adding disk.refresh.help to prototype file 
        SOL.
(80074) Final checkin of 3.2 localized man pages
(80698) vxrelocd not sending e-mail in language of locale, only English.
(80913) vxconfigd dmp_make_mpnode() can unncessarily creates /dev/vx/*dmp nodes
(81083) FMR:panic from snapprint in 3.2
(81274) I18N: Missing exported variable in msg 1001 in disk.ckinit.sh
(81526) Missing a page mapping, causes failed reads in > 2.4.7 kernels
(81622) I18N: minor error in vxrelocd, where vxprint header output was not I18N'ized
(81803) Create new vxdmp tunable "dmp_pathswitch_blocks_shift" to control amount of 
        IOs into dmp path
(35074) vxdg deport/destroy allows you to deport DG with BCV attachments
(42013) EMC: New device types in SYMAPI should be handled properly in libarray
(43079) EMC: vxdg doesn't apply EMC rules when dgid specified, instead of dg name
(43263) General updates and corrections to VM man pages
(54372) Sol: Core dump by vxconfigd when paths to all disks in dg are removed
(56568) EMC: dgrules ignore no mem error condition.
(58021) EMC: Incorrect use of vxsym_errno in vxvm
(58257) vxio and dmp drivers must fail insane disks without retrying i/o to them
(60036) Change vxdisk to implement disk array sepecific rules
(61141) I18N: Need appropiate error handling when user input non-ascii characters
4414884 (62449) Free photon license is not enough for a failed drive in A5X00
(64309) malloc() has to be changed to xalloc() in disc_api.c
(66317) CVM: vold hits assert if node aborts during master takeover
(66733) FMR2: new objects need -e expresions in vxprint
(67433) EMC: TCs in vxtf directory need to be changed after fix for 65056
(70126) after 2 node of a 8 node vcs cluster panic, none of the rest node pick up 
        CVM master
(70172) VVR  primary node hung in a 3 primaries/1 secondary configuration
(70228) VVR: Secondary (32 bit debug) panic while running scenario 7 test 3.13
(70413) add SIG licensing to VxVM
(70569) It would be nice to remove ksh dependency from vxcppart.sh
(71362) vxnotify does not send event notification at time dmp path os disconnected
(71748) cfs glm hits assert vxg_api_unregister:3
(71754) vxconfigd hangs while waiting for a lock
(72069) vxdarestore.tc Fails on 3.2(07_11) debug version leading rest of the TC's to 
        abort
4466429 (72742) root disk cannot be mirrored with disk of same size
(72828) rc.sysinit patch needs revisiting...
(73159) vmconfig stress test tc's
(73197) minor fixes for SIG licensing
(73292) 8 node reltrain cluster has no master - vxconfigd cores with SIGSEGV
(73448) FMR:Wrong arg used in volassist/snap.c error message
(73679) Hitachi 9200 command device not excluded during discovery
(73726) VM 3.2 - vxdisk scandisks does not bring up the new disk(s) in to the list
(73740) VVR: vxrecover -np dumps core if rvg/rep is present in the dg.
(73860) panic in vol_accum_stats() when running vvrtc's
(74085) Add DDL support for Fujitsu arrays
(74215) TC: init_stub.sh doesn't use TC_VXVM_STUB_DIR
(74220) Cleanup the handling of return code for volcvm_establish()
(74418) Need to use full path to mount command
(74457) raw binding overflow upon volume creation is broken
(74610) FMR:Data corruption with multiple plex attaches on NPFMR enabled volume.
(74860) 3.2s breakup_dbg.tc scripts errors on sol26
(74926) cvm: vxconfigd assertion failure in dg.c:1408 on running ora_sm.tc#6
(75338) FMR2: sio queue corruption in dco error handling path
(75560) RAID-5 Logging Problem
(75606) vxconfigd terminated during its initialization if vxdctl enable issued.
(75646) VVR: panic if all the logs for volume are disabled when dcm is active
(75772) no DA record corresponding to RDAC devices
(75953) CVM:vxdg upgrade command fail to upgrade DG version by default
(76000) Add junk for vmdoc package
(76180) RFE: Improve load balancing in DMP for path failure cases.
(76213) CVM: kmsg has timing window in which sender misses notification
(76390) race between dco log volume detach and plex detach
(76466) VVR: vxrlink pause when SRL is around 99% full behaves incorrectly.
(76477) VVR - stop rvg doesn't check for open volume
(76498) USETYPE of var,usr,opt changed to gen in vxvm 3.1.1p2 after using vxinstall 
        to encapsulate the boot disk
(76522) VVR: Revamp flow control algo & implement missing packet resend
4468856 (76691) VM 3.2 controllers with greater than 8 LUNs not recognized
(76719) inst.custom and inst.quick scripts needs minor verbage changes
(76841) T3 Subpath displays are incorrect after failover
(77017) vxdiskadm option 17 prevent multipathing with VID:PID option dump core
(77164) Running 'vxdmpadm getctlrgrp' dumps core
(77262) I18N: Checkin of localized online doc files (man pages, help, etc) for 3.2
(77266) Vold dumps core in handling of Transaction failure because of no accessible 
        config copies
(77306) I18N: Merging in 3.2 I18N shell-script fixes
(77307) I18N:  Merging in 3.2 I18N C source files changes
(77540) Redundant 2.6 patch info needs to be removed
(77987) No more lite license without a key
(78016) Small change needed to support 2.4.7 kernel
(78153) NEC diskarrays like iStorage 4000 are not recognized correctly in VM3.2
(78325) VVR: possible to have two attached rlinks on secondary
(78384) vxconfigd uses /bin/rm in remove_dir()
(78412) DMP: Provide dmp_getsubpaths API in libvxvmsc.so
(78415) Disabling multipathing for a device may result in duplicate entries in vxdisk
        list output
(78428) cvm dmp: volcvm_setiopath ioctl panics due to zero byte allocation
4490913 (78621) vxddladm accepts ncludearray as argument and does include the lib 
                back
(78679) dogi_darecs_on_device uses incorrect grep expression
4495696 (78746) Can not change drives selection during vxinstall as shown in the 
                install menu
(78831) Solaris vxvm-sysboot.sh needs to assure English locale for vxconfigd in
        Japan etc locales.
(78879) CVM: volumes can be written even if dg is not write activated
(79308) Check in missing help file disk.refresh.help
(79311) vxsparecheck needs to send mail compatible with locales.
(79585) Man page changes requested by Japanese l10n team.
(79763) Changes requested by Japanese localisers to vxinfo man page
(79845) update modinfo and pstamp for VM 3.2t Patch 1
(79909) EMC: support VxTF operations in vxtask
(79959) I18N checkin of help files, and adding disk.refresh.help to prototype file 
        SOL.
4510516 (80310) VxVM 3.2T/Monterey: "New Volume" GUI hangs/freezes during creating 
                volume.
(80487) arrays still connected after removing all the array library
(80599) CVM:abortsio can hang for drl volumes
(79959) I18N checkin of help files, and adding disk.refresh.help to prototype
        file SOL.
(81622) I18N: minor error in vxrelocd, where vxprint header output was not
        I18N'ized
(79959) I18N checkin of help files, and adding disk.refresh.help to prototype
        file SOL.
(80074) Final checkin of 3.2 localized man pages
(80698) vxrelocd not sending e-mail in language of locale, only English.
(81274) I18N: Missing exported variable in msg 1001 in disk.ckinit.sh
(61141) I18N: Need appropiate error handling when user input non-ascii
        characters
(77262) I18N: Checkin of localized online doc files (man pages, help, etc)
        for 3.2
(77306) I18N: Merging in 3.2 I18N shell-script fixes
(77307) I18N:  Merging in 3.2 I18N C source files changes
(77540) Redundant 2.6 patch info needs to be removed
(78708) update modinfo and pstamp for I18N build a
(78831) Solaris vxvm-sysboot.sh needs to assure English locale for vxconfigd
        in Japan etc locales.
(79308) Check in missing help file disk.refresh.help
(79311) vxsparecheck needs to send mail compatible with locales.
(71748) cfs glm hits assert vxg_api_unregister:3
(71754) vxconfigd hangs while waiting for a lock
(73292) 8 node reltrain cluster has no master - vxconfigd cores with SIGSEGV
(74303) CVM 3.2r_spfs vxconfigd cores in krecover_dg_purge during recovery_C.tc
(74610) FMR:Data corruption with multiple plex attaches on NPFMR enabled volume.
(109051) ../kernel/vxdmp/aa/vol_disable_dbg.tc#1 failing on 3.2_p3
(90846) VVR: secondary panics because sio_private field is not getting cleared
(92543) VVR: panic occurs while trying to do a vxvm_delay on the ICS
(95246) VVR: panic during readback processing
(95536) VVR: panic in vol_ru_hdr_valid when disassociating SRL.
(96220) VVR: primary panic due to sleep in Interrupt Context
(99368) VVR: secondary pause hangs due to a deadlock situation
(103127) VVR: data corruption due to primary assuming to have sent data when it
         has not
(103491) VVR: vxconfigd dumps core due to mistake in "msg"
(103845) VVR: panic due to holding of two locks in vol_rvdcm_unlog_update
 
 
(from 112385-02)
 
Rebuilt 112385-01 to fix PSTAMP.
 
 
(from 112385-01)
 
4653867 (73206) vxassist is failed to create RAID-5 and Stripe-pro volume on MPxIO 
                enabled SENA devices when only 98.lic license
 
 
==================================================================
Table of Contents
------------------------------------------------------------------
 Section 1: Installing VxVM 3.2 Patch 5
  1.1 Required Sun Patches
  1.2 Prepare to Install the Patch
  1.3 Running patchadd
  1.4 JBOD Array Considerations
  1.5 Removing the VxVM 3.2 P5 Patch
  1.6 Installing on a Sunfire V880 System
 
 
Section 2: Upgrade_Start/Finish
 
Section 3: New Support Information
  3.1 DMP Support for VERITAS ServPoint SAN software
  3.2 New Support for ATF Version 3.4.0.7
 
Section 4: DMP Tunable: dmp_failed_io_threshold consideration.
  IMPORTANT: Wrong dmp_failed_io_threshold setting can lead to data loss
 
Section 5: Full Disk Mirroring and Replacement Considerations
 
Section 6: Duplicate Disk ID's
 
Section 7: vxddladm listsupport Considerations
 
Section 8: vxassist relayout Considerations
 
Section 9: Using vxunroot to Unencapsulate the Rootdisk
 
Section 10: Using devfs to replace a disk
 
Section 11: T3B Firmware Upgrade on Solaris 9
 
Section 12: Error Message when Booting off an ATF Disk
 
Section 13: Error Message when Starting VxVM in Boot Mode
 
Section 14: Updated Manual Pages
 
Section 15 Using Hitachi Arrays
 
Section 16: Workarounds
  16.1 VMSA CFS mount issue
 
Section 17: Troubleshooting a Duplicate Device Entry in vxdisk list
 
Section 18: Documentation Errata
  18.1 Volume Manager 3.2 Installation Guide Changes
  18.2 Volume Manager 3.2 Troubleshooting Guide Changes
  18.3 Volume Manager 3.2 Hardware Notes Changes
  18.4 Volume Manager 3.2 Administrator's Guide Changes
 
Section 19: Disk Connectivity Policy
  19.1 Global Disk Detach Policy
  19.2 Local Disk Detach Policy
 
Section 20: Backing up VxVM Configuration Data
 
Section 21: Disk Encapsulation - Overlapping Partitions
 
Section 22: Updating jbod.info file on a V880 internal disk enclosure configuration
------------------------------------------------------------------
 
 
================================
Section 1
Installing the VxVM 3.2 P5 Patch
================================
 
NOTE: Unlike the VxVM 3.1.1 patch, this same VxVM 3.2 P5 patch can 
be used on all VxVM 3.2 localized releases.
 
For Solaris 9 releases, refer to the man pages for instructions on
using 'patchadd' and 'patchrm' scripts provided with Solaris.
Any other special or non-generic installation instructions should be
described below as special instructions.  The following example
installs a patch to a standalone machine:
 
        example# patchadd /var/spool/patch/112385-06
 
The following example removes a patch from a standalone system:
 
        example# patchrm 112385-06
 
For additional examples please see the appropriate man pages.
 
NOTE: You need to use the shutdown command to reboot the system after patch
installation or de-installation:
 
shutdown -g0 -y -i6
 
NOTE: Installation of VxVM 3.2 P5 involves stopping/starting
vxconfigd multiple times. If VM is not configured prior to
installation of VxVM 3.2 P5, you might see the follwing message:
 
vxvm:vxconfigd: ERROR: enable failed: Volboot file not loaded
        transactions are disabled.
 
 
1.1 Required Sun Patches
None.
 
1.2 Prepare to Install the Patch
 
NOTE:
If you have not yet installed Volume Manager 3.2, you need 
to follow the sequence of events:
 
1. Run pkgadd 3.2s9
2. Run patchadd (see 1.3, below)
3. Run vxinstall
 
Do NOT run vxinstall before running patchadd.
 
Also, before running patchadd or patchrm, you need to:
   -shutdown any application that is accessing the volumes
   -unmount any mounted volumes.
 
To do this, you:
   # umountall
   # vxdctl stop
   # vxiod -f set 0
 
1.3 Running patchadd
To install the patch, run patchadd:
   # patchadd <patchname>
 
NOTE: patchadd requires a patch name (<patchname>).
 
1.4 JBOD Array Considerations
If you have an array which was previously claimed as jbod,
you may get the following warning message during reboot after the upgrade:
   NOTICE: vxvm:vxdmp: added disk array <disk-array-serial-no>
   vxvm:vxconfigd: WARNING: File /etc/vx/array.info not in proper
   format.
   Regenerating file /etc/vx/array.info.
   All attributes will be set to default values.
   The system is coming up.  Please wait.
 
This message informs you that your array configuration has 
changed and that VxVM is recreating the file for you.
 
1.5 Removing the VxVM 3.2 P5 Patch
 
On backing out the VxVM 3.2 P5 patch, you may experience one or both of 
the following:
 
1. You may get the following warning message when you reboot:
   vxvm:vxconfigd: WARNING: Incomplete serial no in file /etc/vx/array.info
   Attributes will be set to default values for these records.
 
NOTE:
   The above warning message may be repeated many times. It can safely
be ignored.
 
2. vxconfigd may core dump upon reboot after removing a patch using patchrm. 
(See TechNote ID: 242565 and Incident 84269) 
 
(The core dump can be removed, it has no effect on how VxVM runs on the 
system.)
 
The core was caused by the old vxconfigd (pre-patch version). Since the core 
points to thread failure, it is likely to be a thread library issue. By doing a 
patchrm, the old libthread.so.1 is not removed, causing the problem.
 
 
1.6 Installing on a SunFire V880 System
Do not install VxVM 3.2 patch 05 if you only have a Photon license
running on a SunFire V880 system.
Customers with V880 systems should insure that they have a properly
installed full VxVM license before installing 3.2 patch 05.
 
Prior to VxVM 3.2 patch03, a photon license will be improperly granted to
and used by the V880 internal array. With the patch installed, a full VxVM
license must be present for VxVM to function on the internal array.
If the root disk is encapsulated, the system should still boot to allow
installation of the license but no VxVM operations will be allowed on the
internal array. If an external photon is attached, the photon license will
still appear but it will not allow VxVM to function on the internal array.
 
The license can be verified using the vxlicense command.
 
PHOTON license:
# vxlicense -p
vrts:vxlicense: INFO: Feature name: PHOTON [98]
vrts:vxlicense: INFO: Number of licenses: 99
vrts:vxlicense: INFO: Expiration date: No expiration date
vrts:vxlicense: INFO: Release Level: 20
vrts:vxlicense: INFO: Machine Class: 200513877
 
For a full VxVM license,  vxlicense -p will return the feature name
"VxVM [95]".
 
 
------------------------------------------------------------------
 
================================
Section 2
Upgrade_Start/Finish
====================
 
You cannot perform upgrade_start/upgrade_finish with the 3.2 
VRTSvxvm package on Solaris 9.
 
------------------------------------------------------------------
 
=======================
Section 3
New Support Information
=======================
 
3.1 DMP Support for VERITAS ServPoint SAN software
The ServPoint SAN product can be connected to servers or platforms supported
by VxVM 3.2 or higher, and it will appear to that server or platform
as a standard supported disk array.
 
3.2 New Support for ATF Version 3.4.0.7
ATF version 3.4.0.7 is now supported.
 
------------------------------------------------------------------
 
===================================================
Section 4
DMP Tunable: dmp_failed_io_threshold consideration.
(Ref incident 136128)
===================================================
 
SYMPTOM:
Possible data loss after failure of a single path in a multipathed 
environment can occur due to interactions between an Application or 
File System, and Volume Manager.
 
This issue only applies to configurations where all the following
apply:
 
1)  Volume Manager version 3.2 P2 (or later) is installed.
2)  Volume Manager Volumes are NOT mirrored.
3)  There are multiple paths to the disk.
4)  Some portion of the I/O path below DMP fails in such a way that 
the error is not returned to DMP before 10 minutes have elapsed. 
Examples of situations that have been seen to cause this include: Bad 
hardware (HBA, FCOT, GBIC, Switch Port, Array Controller), and HBA 
configured to retry endlessly.
 
NOTE:  It is the failure condition in 4) that triggers the problem.  
These failures are rare and are not seen during normal operation of a 
healthy SAN.
 
BACKGROUND:
With Volume Manager 3.2p2, a new tunable was introduced that allows
DMP to explicitly mark a device as "insane" whenever a specific time
period is reached. Normally, DMP would retry on all available DMP 
paths before it concludes that a disk is "insane." With this parameter, 
it can make this conclusion even before it had a chance to retry all 
paths. This tunable sets a threshold value of 600 seconds (10 minutes) 
for an I/O error to be returned from the device. If the I/O takes longer 
than 10 minutes to return with an error, then DMP assumes that the 
device is insane and passes the error to Volume Manager without 
retrying. This allows Volume Manager to use a mirror of the data to 
satisfy the request without further delay.
 
NOTE: This is not a timeout. No DMP activity will occur after the 10 
minutes passes. DMP only checks the elapsed time of the I/O after it 
is returned by the lower layer. If the elapsed time is greater than 
dmp_failed_io_threshold seconds (default 600) then the error will be 
returned to Volume Manager without retries. DMP will wait as long as 
it takes for the I/O to be returned.
 
However, if the delay in returning the I/O is caused by something in 
the I/O path to the device rather than the device itself, DMP will 
incorrectly return the error to the Volume Manager layer rather than 
retrying the I/O on another path. If the Volume is not mirrored, then 
the error will be passed to the File System or Application layer. 
This can result in the File system marking inodes for deletion when 
they are still valid. If raw volumes are in use, then the application 
could believe that the data on disk is corrupted when it is not.
 
If the Volume is mirrored, Volume Manager will satisfy the I/O from 
the other plex and detach the plex that failed, preventing the volume 
from becoming hung.
 
SOLUTION:
To prevent this possibility in situations where mirrored volumes are 
not used, the threshold should be tuned to a sufficiently high value 
so that it will not be reached (16 hours in this example).
 
To change dmp_failed_io_threshold value, you need to add (or modify) 
the value in /kernel/drv/vxdmp.conf: 
 
    dmp_failed_io_threshold=57600
 
Then reboot.
 
In situations where mirrored volumes are in use and an application 
timeout is being hit when there is still a valid plex with the data, 
the dmp_failed_io_threshold can be tuned to a smaller value to allow 
the I/O to succeed on the mirror without triggering an application 
failure.
 
 
------------------------------------------------------------------
 
==================================================
Section 5
Full Disk Mirroring and Replacement Encapsulated Root Disk Considerations
=========================================================================
 
If you are mirroring or replacing an encapsulated root disk, then it is
strongly recommended that you use vxdiskadm option #6 for mirroring and
option #5 for replacing.
 
However, if you choose to use vxassist to mirror your encapsulated
root disk, or if you wish to manually replace the disk, then you will
need to make sure that your target disk is setup correctly first.
 
To allow for future forwards and backwards compatibility,  a 1-cylinder 
gap was introduced between the private and public regions 
of the disk in VxVM 3.2; effectively reducing the public region by one
cylinder. Manual instruction is now needed to create a disk layout on
the target disk first. 
 
Consider these two scenarios :   
1. Mirroring a full disk created using an earlier version of VxVM 
   to a target disk which is exactly the same size as the 
   source disk. (A "full disk" is a disk that has no free cylinders.)   
   NOTE: You only need to use this step if disk mirroring/replacement
   using vxdiskadm fails. For details, see vxdiskadm(1M).   
2. Replacing a failed disk created using an earlier version of 
    VxVM to a target disk which is exactly the same size as the 
    source disk.
 
NOTE: 
   In the following examples, the traditional c#t#d# naming
   scheme is used. If you are using enclosure-based naming scheme,
   use the enclosure name instead. 
 
1. Determine the size of the source disk's private region, 
   using either of the following methods:   
   a. Get the private region length of the source disk using:
	#vxdisk list diskname
    where diskname is the disk media name of the source/failed disk. 
    In the displayed output, note the len(length) value for 
    the private field.
    In this example, the value of this field is 3071: 
	private: slice=4 offset=1 len=3071 
   b. For failed disk, if this information is not be available, 
    you must decide what the private region is. 
    (One option is to use the same private region size of another 
    identical disk in the disk group.)  
 
2. Use the vxdisksetup program to initialize the target disk: 
    # /usr/lib/vxvm/bin/vxdisksetup -i c#t#d# privlen=XXXX old_layout
   where XXXX is the size of the source disk's private region. 
 
3. Add the newly initialized target disk to the source disk group: 
	# vxdg  -k -g diskgroup adddisk medianame=c#t#d# 
 
------------------------------------------------------------------
 
===================
Section 6
Duplicate Disk Id's
===================
When VxVM detects disks with duplicate disk IDs, VxVM attempts to 
select the appropriate disk (using array vendor-specific logic). If a 
disk cannot be selected, VxVM does not import any of the duplicated 
disks into a disk group. 
In the rare case when VxVM cannot make the selection, you must choose 
which duplicate disk to use. (An array with hardware mirroring capability 
is particularly susceptible to data corruption, if the wrong disk were to
be selected.)
The following examples describe situations where user intervention is 
needed.
 
Example 1:
When DMP is disabled to an array that has multiple paths, then each path 
to the array is claimed as a unique disk. VxVM detects duplicate disks 
whenever an array is configured as a unique disk for each accessible 
path and gives the following message:
vxvm:vxconfigd: NOTICE: Unable to resolve duplicate diskid.
When DMP is suppressed, VxVM can not know which path to select as the 
true path. You must decide which path to use. Decide which path to exclude
 and then either edit the file /etc/vx/vxvm.exclude or, if vxconfigd is 
running, use the vxdiskadm option 17 selection 1 (suppress all paths 
through a controller from VxVM?s view) or selection 2 (suppress a path 
from VxVM?s view).
The following example shows a vxvm.exclude file with paths c6t0d0s2, 
c6t0d1s2, and c6t0d2s2 excluded from VxVM:
exclude_all 0
paths
c6t0d0s2 /pci@1f,4000/SUNW,ifp@2/ssd@w50060e8003275705,0
c6t0d1s2 /pci@1f,4000/SUNW,ifp@2/ssd@w50060e8003275705,1
c6t0d2s2 /pci@1f,4000/SUNW,ifp@2/ssd@w50060e8003275705,2
#
controllers
#
product
#
pathgroups
 
Example 2:
Some arrays such as EMC, HDS, and so on provide hardware mirroring. 
When a lun pair is split, depending on how the process is performed, 
you may get two disks with the same diskid. With duplicate disk IDs, 
VxVM gives the following message:
vxvm:vxconfigd: NOTICE: Unable to resolve duplicate diskid.
Check with your array vendor to make sure that your site uses the 
proper split procedure. If you know which luns to use, decide which 
path to exclude and then either edit the file /etc/vx/vxvm.exclude or,
 if vxconfigd is running, use the vxdiskadm option 17 selection 1 
(suppress all paths through a controller from VxVM?s view) or 
selection 2 (suppress a path from VxVM?s view).
 
Example 3:
When you have disks duplicated using dd or any other disk copying 
utility, VxVM gives the following message:
vxvm:vxconfigd: NOTICE: Unable to resolve duplicate diskid.
Choose which set of duplicated disks that you want to exclude, and 
either edit the file /etc/vx/vxvm.exclude or, if vxconfigd is running,
 use the vxdiskadm option 17 selection 1 (suppress all paths through 
a controller from VxVM?s view) or selection 2 (suppress a path from 
VxVM?s view).
 
 
------------------------------------------------------------------
 
===================================
Section 7 
vxddladm listsupport Considerations
===================================
After installing your patch, you may not see all the supported arrays 
when using the vxddladm listsupport command.
If you do not see the correct information, enter:
   # vxddladm addsupport all
   # vxddladm listsupport
 
to see updated results.
 
NOTE: A core dump may occur when using the vxddladm addsupport all command
(refer to Sun BugID 4642144). Until this problem is fixed, a temporary 
work-around is to set the LD_LIBRARY environment variable to /usr/lib 
before running vxddladm addsupport all.
------------------------------------------------------------------
===================================
Section 8
vxassist relayout Considerations
===================================
 
The vxassist relayout operation requires all mirrors in the volume
to have the same layout. If the volume is contains mirrors with
different layouts, then you need to relayout the mirror plexes
to the same layout before performing the volume relayout operation.
------------------------------------------------------------------
 
============================================
Section 9 
Using vxunroot to unencapsulate the root disk
=============================================
Only those encapsulated volumes that were derived from the pre-encapsulated 
root disk partitions can be unrooted. All other volumes created on the root 
disk after encapsulation must be either removed or moved to another disk
before using vxunroot.
 
------------------------------------------------------------------
 
========================================
Section 10
Using devfs to replace a disk
========================================
 
Before using devfs to replace a disk, you should ensure that the dev_t
entry for the old disk is removed from the OS device tree. Use the following
command to remove the dev_t entry:
 
luxadm -e offline <device_path>
 
----------------------------------------------------------------------
=================================
Section 11
T3B Firmware Upgrade on Solaris 9
=================================
On Solaris 9 only, a T3B upgrade to firmware version 2.1 must follow
the procedure below. Not using the procedure leads to disabled disk
groups or an inability to mount file systems. The procedure is a result
of VERITAS incident number 95877.
Use the mount command to unmount related filesystems.
 
Stop all VxVM volumes:
# vxvol stop <vol_name>
 
Stop VxVM:
# vxdctl stop
#vxiod -f set 0
 
Upgrade T3B firmware to version 2.1.
 
Start VxVM:
# vxiod set 10
# vxconfigd -m disable
# vxdctl enable
Start the VxVM volumes:
 
# vxvol -g <dg_name> start <vol_name>
Use the mount command to remount the file system.
----------------------------------------------------------------
==============================================
Section 12
Error Message when Booting off a Target Driver
==============================================
 
An updated When any target driver, such as an ATF, coexists with VxVM,
and if the target driver happens to claim the boot disk, then the
following message may be displayed at the console:
WARNING: vxvm:vxdmp: Cannot find device number for </pci@1e,2000/lpfc@2/sd@0,7:a> 
where the path </pci@1e,2000/lpfc@2/sd@0,7:a> corresponds to the bootdisk,
obtained from EEPROM.
This message can safely be ignored. (Ref. incident 109019.)
---------------------------------------------------------------------------------------------
 
=============================================
Section 13
Error Message when Starting VxVM in Boot Mode
(Ref. incident 102494)
=============================================
In some situations, a licensing check may display the following error message
when starting VxVM in boot mode: 
 
/usr/sbin/prtconf: getexecname() failed
 
This message can be safely ignored. (Refer also to Sun Bug# 4122428 and 4209527.)
 
Workaround:
 
#
# Find the real hardware architecture:
#
 
prtconf | head -5 | tail -1
 
cd /platform
ln -s sun4u SUNW,Ultra-250
cd /usr/platform
ln -s sun4u SUNW,Ultra-250
sync
reboot
 
-------------------------------------------------------------------------------------
 
====================
Section 14
Updated Manual Pages
====================
 
You should replace the manual pages issued with the previous VxVM 3.2 release with
new manual pages. Please refer to TechFile ID 251542 (at

http://support.veritas.com/docs/251542) for instructions on how to obtain the latest
manual pages package, VRTSvmman.
 
------------------------------------------------------------------
 
======================
Section 15
Using Hitachi Arrays
(Ref. incident 100458)
======================
If you are considering having your boot disk residing on an Hitachi array
you should first contact Hitachi Data Systems for the latest information
on supported boot disk configurations with Hitachi arrays.
------------------------------------------------------------------
 
==========================
Section 16
Workarounds
==========================
 
16.1 VMSA CFS mount issue: Workaround 
 
Problem:
 
VMSA fails to list/recognize valid cluster nodes 
for a CFS mount operation, when the CFS is not
mounted on at least one cluster nodes.
 
Symptom:
 
When attempting a cluster-mount operation on an
unmounted CFS: 
 
VMSA rejects valid cluster node names in mount 
dialogs.
 
VMSA fails to list cluster nodes in dialogs that
provide a list of valid nodes for mounting a CFS.
 
Workaround:
 
Execute the following cluster-mount command on
the cluster's primary node:
 
# /opt/VRTSvxfs/lib/vcs/mount_cfs /<block_device_name> <node_name>
 
where <block_device_name> is the block-device name for the
CFS, and <node_name> is the name of any node in the cluster.
 
After mounting the CFS on one cluster node, VMSA will 
function normally.
------------------------------------------------------------------
 
=======================================================
Section 17
Troubleshooting a Duplicate Device Entry in vxdisk list
(Ref. incidents 114479 and 101371)
=======================================================
 
Please follow this procedure if you encounter duplicate entry in
"vxdisk list" or vxdisksetup gives "Duplicate DA" error as
specified in Incident:101371 SunBug:4630477 and Incident:114479
SunBug:4769704.
 
# vxdisksetup -i c1t5d0
vxdisksetup: c1t5d0: Duplicate DA records encountered for this device.
                     Refer to the troubleshooting guide to clear them
 
# vxdisk list
<snip>
c0t8d0s2     sliced    -            -            online
c1t2d0s2     sliced    c1t2d0s2     rootdg       online
c1t3d0s2     sliced    c1t3d0s2     ttdg         online
c1t5d0s2     sliced    c1t5d0s2     -         error
c1t5d0s2     sliced    c1t5d0s2     -         online/error <<<-- any
state is ok.
<snip>
 
1. Remove c1t5d0s2 entries from vxvm control.
 
   Do "vxdisk rm <da-name>" for all the duplicate entries
 
   [since you don't know which one is the valid one, so do
   it for all ...there can be more than 2 duplicate entries
   also ...]
 
 vxdisk rm c1t5d0s2
 vxdisk rm c1t5d0s2 <-- do it again to remove all the entries.
 
 
2. Remove the disk c1t5d0s2 using luxadm ( solaris command ).
 
   Get A5K Array name and Slot number of the disk using luxadm
 
   #luxadm disp /dev/rdsk/c1t5d0s2
 
 
   Remove device c1t5d0s2 using "luxadm remove_device" command.
   [luxadm remove_device enclr,slot-number]
 
   #luxadm remove_device SAHYADRI,f5
 
 
   Pull the disk out as per luxadm instructions.
 
 
3. Run command  "devfsadm -C"
 
4. Run command  "vxdctl enable"
 
 
  [Till this point, we have removed the dev_t corresponding to the
   physical disk. Now we will remove all the stale dev_t's]
 
  Please loop ac per below instructions.
 
LOOP :
 
  [you will see one entry less, since we can have more than 2 duplicate
   entries]
 
  #vxdisk list
 <snip>
 c0t8d0s2     sliced    -            -            online
 c1t2d0s2     sliced    c1t2d0s2     rootdg       online
 c1t3d0s2     sliced    c1t3d0s2     ttdg         online
 c1t5d0s2     sliced    c1t5d0s2     -         error
 <snip>
 
5. Again remove *ALL* duplicate c1t5d0s2 entries from vxvm control.
 
 vxdisk rm c1t5d0s2
 
6. Run command "luxadm -e offline <path to disk>" on *ALL THE PATHS*
   to the disk [this removes the stale dev_t]
 
   Our test machine is having 2 paths to the disk through
    controllers c1 and c2
 
 #luxadm -e offline /dev/dsk/c1t5d0s2
 #luxadm -e offline /dev/dsk/c2t5d0s2
 
7. Run command "devfsadm -C"
 
8. Run command  "vxdctl enable"
 
goto LOOP:
 
[Continue this process until there are no more entries in vxdisk list
 of corresponding disk c1t5d0s2 ]
 
Result:
 
# vxdisk list
<snip>
c0t8d0s2     sliced    -            -            online
c1t2d0s2     sliced    c1t2d0s2     rootdg       online
c1t3d0s2     sliced    c1t3d0s2     ttdg         online
<snip>
 
 
Now both OS device tree and VxVM are in a clean state corresponding
to disk c1t5d0s2.
 
Please follow the procedure to replace the failed disk or removed disk
as described in the Volume Manager Administration Guide (replace a new disk in
place of device c1t5d0s2).
 
 
==================================
Section 18
Documentation Errata for VxVM 3.2
==================================
 
Please note the following changes to the documentation 
for VxVM 3.2 for Solaris.
 
-------------------------------------------------
18.1 VERITAS Volume Manager Installation Guide
-------------------------------------------------
[1] On page 9, reads:
 
The boot disk is a special case. If no other space is
available, VxVM attempts to allocate space usually reserved
for swap by shrinking the swap partition (which must be slice
1). This process is known as swap relocation and, if
necessary, happens automatically during root disk
encapsulation.
 
[Should be]
 
The boot disk is a special case. If no other space is
available, VxVM attempts to allocate space usually reserved
for swap by shrinking the swap partition. This process is
known as swap relocation and, if necessary, happens
automatically during root disk encapsulation.
 
** Notice the phrase "(which must be slice 1)" is removed
 
 
[2] On page 53, Step 2 reads:
Determine if the root disk is encapsulated by running the
following command:
# df -k
 
If the output from this command includes a pathname that contains
vx and rootvol as in /dev/vx/dsk/rootvol, then the rootdisk is 
encapsulated and you must upgrade both VxVM and Solaris. See 
"Upgrading VxVM on an Encapsulated Root Disk on page 56."
 
[Should be]
If the output from this command includes a pathname that contains
vx and rootvol as in /dev/vx/dsk/rootvol, then the rootdisk is encapsulated. 
See "Upgrading VxVM on an Encapsulated Root Disk
on page 56."
 
-------------------------------------------------
18.2 VERITAS Volume Manager Troubleshooting Guide
-------------------------------------------------
[1] On page 33
rm -r /etc/vx/reconfig.d/state.d/installdb
 
[should be]
rm -r /etc/vx/reconfig.d/state.d/install-db
 
[2] On page 47
or invoke vxconfigd as shown here:
 
[should be]
or invoke vxconfigd under C locale as shown here:
 
[3] On page 48
specify the option -x syslog to vxconfigd:
 
[should be]
specify the option -x syslog to vxconfigd under C locale:
 
-------------------------------------------------
18.3 VERITAS Volume Manager Hardware Notes
-------------------------------------------------
[1] On page 5
path c1t0d0s2 is represented by the DMP metanode c1t0d0s2 
which isrepresented by device nodes in the /dev/vx/dmp and
/dev/vx/rdmp directories.
 
[should be]
path c1t0d0s2 is represented by the DMP metanode c2t0d0s2 
which is represented by device nodes in the /dev/vx/dmp and
/dev/vx/rdmp directories.
 
[2] On page 5
This example output shows that the path c2t0d0s2 is owned 
by the DMP device c2t0d0s2 which has two paths to it.
 
[should be]
This example output shows that the physical path c1t0d0s2 
is owned by the DMP device c2t0d0s2 which has two paths to it.
 
[3] On pages 16 and 17
In the "DMP and JBOD Disk Array Support," section, on pages 16-17, 
the instructions for adding JBOD disk arrays to the JBOD category
of DMP by editing vxdmp.conf are no longer supported, and should 
be ignored. You must use the vxddladm addjbod command to add disks 
to the JBOD category.
Please refer to the vxddladm(1M) man page for details.
 
[4] On page 23
#vxdmpadm getsubpaths ctlr=/c1
 
[should be]
#vxdmpadm getsubpaths ctlr=c1
 
[5] On page 23
NAME STATE TYPE DMPNODENAME DA-TYPE DA-SNO
 
[should be]
NAME STATE TYPE DMPNODENAME ENCLR-TYPE ENCLR-NAME
 
[6] On page 29
/dev/RAID_module01/0s[0-7] and /dev/RAID_module01/0s[0-7]
 
[should be]
/dev/RAID_module01/0s[0-7] and /dev/rRAID_module01/0s[0-7]
 
[7] on page 34
Replace the complete section under "Sun StorEdge T3 Support in A 
Multi-host Environment"
 
[should be]
In multi-host configurations, the Sun StorEdge T3 must be configured 
to the MPxIO mode (explicit failover or nonauto_trespass). In this 
mode, the Sun StorEdge Traffic Manager Software (STMS) driver on the 
host handles multipathing. VERITAS Volume Manager (VxVM) will not 
see multiple paths to any device on the array.
 
You need to upgrade the Sun StorEdge T3 firmware to 1.17B. The patch 
for this upgrade is 109115-09.
 
The /kernel/drv/scsi_vhci.conf file must have the MPxIO mode enabled 
(mpxio-disable="no"). Use the command "sys mp_support mpxio" on the 
Sun StorEdge T3 array to enable the MPxIO mode.
 
NOTE: All hosts in the clustered environment must use the MPxIO mode.
 
 
-------------------------------------------------
18.4 VERITAS Volume Manager Administrator's Guide
-------------------------------------------------
[1] On page 46
"The original volume must not be in use during a snapback
operation in which it is resynchronized from a snapshot."
 
[should be]
The original volume must not be in use during a snapback 
operation that uses the option -o resyncfromreplica to 
resynchronize the volume from the snapshot.
 
[2] On page 110
ENCLR_NAME ENCLR_TYPE ENCLR_SNO              MODE        STATUS
==================================================================
enc0       T300       60020f20000001a90000   PRIVATE     CONNECTED
 
[should be]
 
ENCLR_NAME ENCLR_TYPE ENCLR_SNO              STATUS
=======================================================
enc0       T300       60020f20000001a90000   CONNECTED
 
[3] On page 111
ENCLR_NAME ENCLR_TYPE ENCLR_SNO              MODE       STATUS
=================================================================
others0    OTHER      OTHER_DISKS            PRIVATE    CONNECTED
seagate0   SEAGATE    SEAGATE_DISKS          PRIVATE    CONNECTED
enc0       T300       60020f20000001a90000   PRIVATE    CONNECTED
 
[should be]
 
ENCLR_NAME ENCLR_TYPE ENCLR_SNO              STATUS
=========================================================
others0    OTHER      OTHER_DISKS            CONNECTED
seagate0   SEAGATE    SEAGATE_DISKS          CONNECTED
enc0       T300       60020f20000001a90000   CONNECTED
 
[4] On page 111
ENCLR_NAME ENCLR_TYPE ENCLR_SNO              MODE       STATUS
=================================================================
others0    OTHER      OTHER_DISKS            PRIVATE    CONNECTED
seagate0   SEAGATE    SEAGATE_DISKS          PRIVATE    CONNECTED
GRP_1      T300       60020f20000001a90000   PRIVATE    CONNECTED
 
[should be]
 
ENCLR_NAME ENCLR_TYPE ENCLR_SNO              STATUS
=======================================================
others0    OTHER      OTHER_DISKS            CONNECTED
seagate0   SEAGATE    SEAGATE_DISKS          CONNECTED
GRP_1      T300       60020f20000001a90000   CONNECTED
 
 
[5] On page 155
# vxmake -hfsgen vol home plex=home-1,home-2
 
[should be]
 
# vxmake -U fsgen vol home plex=home-1,home-2
 
[6]on page 220
or specify the start keyword to vxrelayout, as shown here:
# vxrelayout -o bg start vol04
 
[should be]
For relayout operations that have not been stopped using the 
vxtask pause command (for example, the vxtask abort command 
was used to stop the task, the transformation process died, 
or there was an I/O failure), resume the relayout by specifying 
the start keyword to vxrelayout, as shown here:
 
# vxrelayout -o bg start vol04
 
NOTE. If you use the vxrelayout start command to restart a relayout 
that you previously suspended using the vxtask pause command, a new
untagged task is created to complete the operation. You cannot then
use the original task tag to control the relayout.
 
[7] On page 272
 
9. On the OHP host where the backup is to be performed, 
use the following command to import the snapshot volumes
fs disk group:
# vxdg import snapvoldg
 
[should be]
 
9. On the OHP host where the backup is to be performed, 
use the following command to import the snapshot volumes
fs disk group:
# vxdg import snapvoldg
The snapshot volume is initially disabled following the split.
Use either of the following commands on the OHP host to 
restart the snapshot volume:
# vxrecover -g snapvoldg -m snapvol
# vxvol startall
 
 
[8] On page 2
 
10. On the OHP host where the replica database is to be set up,
use the following command to import the snapshot volumes
fs disk group:
# vxdg import snapvoldg
 
[should be]
 
10. On the OHP host where the replica database is to be set up,
use the following command to import the snapshot volumes
fs disk group:
# vxdg import snapvoldg
The snapshot volume is initially disabled following the split.
Use either of the following commands on the OHP host to restart
the snapshot volume:
# vxrecover -g volumedg -m snapvol
# vxvol startall
 
 
[9] On page 291
Replace the complete section under "voldrl_max_drtregs."
 
[should be]
voldrl_max_drtregs
 
This tunable specifies the maximum number of dirty regions that can exist
on the system at any time. This is a global value applied to the entire
system, regardless of how many active volumes the system has.
 
The default value for this tunable is 2048.
 
The tunable voldrl_max_drtregs can be used to regulate the worse-case
recovery time for the system following a failure.  A larger value may
result in improved system performance at the expense of recovery time.
 
 
========================
Section 19
Disk Connectivity Policy
(Ref incident 118065)
========================
This section provides information relating to BugID 4817095.
 
With VERITAS Volume Manager (VxVM), it is possible to create a shared
disk group on the master node of a cluster. This provides all nodes
in the cluster with concurrent read and write access to the volumes within
the shared disk group.
 
Only the master node can create a shared disk group. This has the following
advantages and implications:
 
* All the nodes in the cluster see exactly the same configuration.
* Only the master node can change the configuration.
* Any changes on the master node are coordinated and propagated to the other
(slave) nodes in the cluster.
* Any failures requiring a configuration change must be sent to the master
so that they can be correctly resolved.
* As the master node resolves any failure, all the slave nodes are correctly
updated; ensuring that all nodes have the same view of the configuration.
 
The practical implication of this approach is that any IO failure on any
node results in the configuration of all nodes being changed.
 
However, in some cases, it is not desirable to have all nodes reacting in this
way to an IO failure (this is known as the global disk detach policy).
To address these cases, an alternative way of responding to IO failures was
added for shared disk groups. Starting with VxVM 3.2 the local disk detach
policy (more formally known as the connectivity policy) became available for
disk groups version 70 and above.
 
NOTE: When the master causes an IO failure on a disk, the disk is marked as
failed and removed from the disk group. This is the same for both local detach
policy and global detach policy. On all the nodes, the volumes on that disk
will be unable to perform IO to an underlying physical device. When a slave
causes an IO failure the disk is not removed from the disk group.
 
19.1 Global Disk Detach Policy
 
The global disk detach policy is the traditional and default policy for all
nodes on the configuration.
 
In this case, if there is an IO failure on one node, the master node
performs the normal IO recovery work to repair the failure, and the plex
is detached cluster wide. All nodes in the cluster continue to perform
IO functions.
 
 
NOTE: The global detach policy must be used with Active/Passive arrays, This
ensures that all nodes correctly coordinate use of the active path.
 
19.2 Local Disk Detach Policy
 
The local disk detach policy was first designed for failover applications in
large clusters.  If an application is affected by IO failures, the cluster
framework is then able to move the application to a node that still had access
to the volume.
The local disk detach policy is used particularly with mirrored volumes.  (For
Unmirrored (or hardware mirrored) volumes there is no difference between
the local and global detach policies.)
 
In the case of local disk detach policy, if there is a write failure on a node,
the master node performs the normal IO recovery work to repair the failure; but
in addition, all the nodes are contacted to see if the disk is still acceptable.
 
If the failure is not seen by all nodes, the local detach policy
stops IO only from the node that had the failure.
 
NOTE: For private disk groups, the local disk detach policy does not change the
behavior of the disk group.
 
==================================
Section 20
Backing up VxVM Configuration Data
==================================
This section addresses BugID 4233507.
 
A new feature, the vxconfigbackupd daemon, is now available for backing up VxVM 
configuration data. 
 
vxconfigbackupd monitors changes to disk group configuration in VxVM. Whenever a 
configuration change occurs, the daemon backs up the changed disk group configuration 
to a file using the vxmake description format.
 
The vxconfigbackupd daemon is started from the /etc/rc2.d/S95vxvm-recover startup 
script. Changes that you make in this script for the invocation of vxconfigbackupd
do not take effect until VxVM is next started from the script.
 
Error and warning messages for the daemon are written with a timestamp to the
console device. The daemon exits on receiving a TERM signal.
 
By default, configuration backup files are stored in the /etc/vx/dgcfg directory in 
compressed form (using gzip or compress), and are maintained in a cyclical manner.
An alternate configuration backup directory can be specified using the -d option.
The files are stored uncompressed if the compression utilities are not available on 
the system. The gzip utility is preferred over compress when both utilities are 
available.
 
The configuration copy files are named dgname.N, where N is a number ranging from 
1 for the most recent copy to 5 for the oldest copy.
The number of copies that are maintained can be changed by using the
-n option.
 
When a disk group is deported or destroyed, the default behavior is to move
the configuration files for the disk group from the backup directory to
the deport subdirectory.
If the -R option is specified, the configuration files are deleted instead of moved.
 
By default, the daemon logs the configurations of all disk groups that are currently 
imported on the system.
If a list of one or more disk group names is specified as an argument,
this restricts the daemon to logging the configurations of only these disk
groups.
 
Please refer to the configbackupd(1M) manual page for further information.
 
===========================================
Section 21
Disk Encapsulation - Overlapping Partitions
(Ref incident 114288)
===========================================
This section provides additional information in relation to BugID 4711196.
When Solaris disks are encapsulated, provision is made for overlapping partitions - 
allowing the public region to overlap the private region. This is action is necessary
because Solaris partitions must be cylinder aligned and offsets must protect the VTOC. 
 
 
=======================================================================
Section 22
Updating jbod.info file on a V880 internal disk enclosure configuration
(Ref incident 133579, Sun BugID 4900508 )
=======================================================================
 
There is a new procedure for supporting multipathed JBOD (Just a Bunch
of Disks) drives with VERITAS Volume Manager (VxVM) 3.2 and later. This
new procedure is needed for multipathed internal disk drives on
Enterprise 3x00 and Sun Fire V880 systems. Customers with internal
drives with VxVM 3.2 cannot enable Dynamic Multipathing (DMP)
support without this procedure, which will lead to serviceability
issues.   
 
This procedure only affects multipathed disk drives that are not
supported under VERITAS Device Discovery Libraries (DDL) in VxVM 3.2.
 
The following is an example of failing configuration:
 
  VERITAS Volume Manager (VxVM)
 
  . VERITAS Volume Manager 3.2 and later
 
  Sun Workstation/Server with internal multipathed disks, example:
 
  . Sun Enterprise Server 3x00
  . Sun Fire V880
 
Sun StorEdge Arrays are supported with VxVM 3.2 DDL and do not need this
procedure for DMP to work. Specifically, Photon (A5x00), Purple I/II
(T3/T3+/T3B), and Sonoma (A3500) are supported with storage specific DDL
and are not affected.
 
The following example shows a configuration which will experience this 
problem.
 
If a multipathed disk is connected to two different HBAs such as c0 and 
c1, 'vxdisk' will list it as two separate disks:
 
  # vxdisk list
 
    --------------------------------------------------
   | DEVICE    | TYPE    | DISK    | GROUP   | STATUS |
   |===========+=========+=========+=========+========|
   | c0t0d0s2  | sliced  | disk0   | rootdg  | online |
   | c1t0d0s2  | sliced  | disk1   | rootdg  | online |
    --------------------------------------------------
 
Here two disks showed up instead of one. Any internal JBOD disk that 
has multiple paths to it is subject to this problem.
 
Another example occurs when the vxdmpadm command shows a multipathed 
disk having only one path to it.
 
  # vxdmpadm getdmpnode nodename=c0t0d0s2
 
    -----------------------------------------------------------------------
  | NAME     | STATE    | ENCLR-TYPE  | PATHS  | ENBL | DSBL | ENCLR-NAME |
  |==========+==========+=============+========+======+======+============|
  | c0t0d0s2 | ENABLED  | OTHER_DISKS | 1      | 1    | 0    | OTHER_DISKS|
  | c0t1d0s2 | ENABLED  | OTHER_DISKS | 1      | 1    | 0    | OTHER_DISKS|
   -----------------------------------------------------------------------
 
This is a change of behavior starting in VxVM 3.2 compared to earlier releases.
VxVM 3.2 moved to a new device discovery library (DDL) architecture in
which specific device support (Hitachi, Photon, T3, Sonoma) is moved
from kernel driver to user libraries (under /etc/vx/lib/discovery.d).
The benefit is that new device support and bug fixes can now be done
with minimal disruption to the system.
 
The solution to this problem for VxVM 3.2 and later is to specify JBOD
support with the 'vxddladm' command. This command needs to be executed
once and the information is persistent across reboots.  Preferably, the
command should be run *before* running 'vxinstall' for the first time.
Details for this procedure can be found in the VxVM 3.2
Administrator's Guide, and are repeated below.
 
 
CORRECTIVE ACTION:
 
The following recommendation is provided as a guideline for authorized
Enterprise Services Field Representatives who may encounter the above
problem. 
 
This procedure can also be used to support Vicom Storage Virtualization 
Engine. It is also used in the StorEdge 6xx0 solution. 
 
New service procedure and documentation change are required to address
this VxVM multipathing issue. VERITAS has documented the 'vxddladm'
JBOD support in the VxVM 3.2 Administrator's Guide, page 59,
under the section:
 
     Administering Disks
     --> Discovering and Configuring Newly Added Disk Devices
     ----> Administering the Device Discovery Layer
     ------> Adding Support for Disks in the JBOD Category
 
Note that "vxdmp.conf" is not the right fix for this case as 
"vxddladm" will eventually take over the functionality.
 
  For completeness, please adhere to the following procedure:
 
  1. Find Disk Vendor ID
 
     Execute the following command to correctly identify the vendor
     ID of the internal disk: (must run as root)
 
     /etc/vx/diag.d/vxdmpinq <VxVM device name in /dev/vx/rdmp/>
 
     For example:
 
       # /etc/vx/diag.d/vxdmpinq /dev/rdsk/c0t1d0s2
       Vendor id (VID)  : SEAGATE 
       Product id (PID) : ST318404LSUN18G 
       Revision         : 8507
       Serial Number    : 0025T0LA3H
 
     In the example above, the vendor id is "SEAGATE".
 
 
  2. Add VERITAS Device Discovery Layer support for JBOD.
     (must run as root)
 
       # /usr/sbin/vxddladm addjbod vid=SEAGATE pid=ST31840LSUN18G
       # vxdctl enable
 
     NOTE: If the device is FUJITSU, you must include the length.
           For example, 
      # /usr/sbin/vxddladm addjbod vid=FUJITSU pid=MAG3091LSUN9.0G
           length=10
 
     Use vxddladm command to verify the JBOD disk is now supported:
 
       # /usr/sbin/vxddladm listjbod
 
 
--------------------------------------------------------------------
       | VID     | PID      | Opcode | Page Code | Page Offset | SNO length
|
 
|=========+==========+========+===========+=============+============|
       | SEAGATE | ALL PIDs | 18     | -1        | 36          | 12
|
 
--------------------------------------------------------------------
 
  3. Reboot the system
 
  4. Verify devices are now multipathed.  
 
       # /usr/sbin/vxdmpadm listenclosure all
 
         ---------------------------------------------------------
        | ENCLR_NAME    | ENCLR_TYPE   | ENCLR_SNO    | STATUS    |
        |===============+==============+==============+===========|
        | OTHER_DISKS   | OTHER_DISKS  | OTHER_DISKS  | CONNECTED |
        | Disk          | Disk         | DISKS        | CONNECTED |
         ---------------------------------------------------------
 
  5. Verify disk and DMP paths are recognized.  
 
       # /usr/sbin/vxdisk list
 
         Devices under Disk_0 are multipathed disks.
 
         ------------------------------------------------
        | DEVICE  | TYPE    | DISK     | GROUP  | STATUS |
        |=========+==========+==========+=======+========|
        | Disk_0  | sliced  | disk01   | rootdg | online |
         ------------------------------------------------   
 
       # /usr/sbin/vxdmpadm getdmpnode enclosure=Disk
 
         ------------------------------------------------------------------
       | NAME   | STATE   | ENCLR-TYPE | PATHS | ENBL | DSBL | ENCLR-NAME |
       |========+=========+============+=======+======+======+============|
       | Disk_0 | ENABLED | Disk       | 2     | 2    | 0    | Disk       |
        ------------------------------------------------------------------

=======================================================================
Section 23
Moving a disk device from one physical array slot to another
(Ref incident 137607)
=======================================================================

PROBLEM:

If a disk device is moved (or swapped) from one physical array slot to another, then it is 
possible for two different devices to be incorrectly merged into a single DMP node 
- ultimately leading to possible data corruption.

Note: This situation can occur even after the disks have been re-discovered (for example,
 by means of the devfsadm -C command) and re-scanned (for example, by using the vxdctl enable 
command).

For example; suppose the disk at slot 2 is powered off. The host will not then see any device 
in slot 2. Then:
 
# vxdisk list
DEVICE       TYPE      DISK         GROUP        STATUS
c1t1d0s2     sliced    -            -            error
c1t1d0s4     simple    c1t1d0s4     rootdg       online
c3t1d0s2     sliced    -            -            online
c3t3d0s2     sliced    -            -            online
c3t4d0s2     sliced    -            -            online
c3t5d0s2     sliced    -            -            online
c3t6d0s2     sliced    -            -            online

Note that c3t2d0 is missing. Also notice that DMP device c3t5d0 has only one entry.

Then:
# vxdmpadm getsubpaths dmpnodename=c3t5d0s2
NAME         STATE     PATH-TYPE  CTLR-NAME  ENCLR-TYPE   ENCLR-NAME
====================================================================
c3t5d0s2     ENABLED     -        c3         Disk         Disk


Suppose, now:
disk c3t5d0 at slot 5 is powered off and removed from slot 5
disk c3t2d0 at slot 2 is removed from slot 2
The two disks are swapped (so that the disk that was in slot 5 is now in slot 2,
and the disk that was in slot 2 is now in slot 5
Both disks are powered on.

If rediscover and rescan are now performed, in an attempt to get DMP and the OS see the 
new disk arrangement, it is possible that DMP node c3t5d0s2 has become "confused," 
and has incorrectly multi-path'ed BOTH the old device (c3t2d0s2) as well as the new device 
(c3t5d0s2) as the same device. Since these two disks are physically different, and should 
never be multi-path'ed into the same DMP node, data corruption will occur if I/O goes 
through this DMP path.
  
# devfsadm -C
# vxdctl enable
# vxdisk list
DEVICE       TYPE      DISK         GROUP        STATUS
c1t1d0s2     sliced    -            -            error
c1t1d0s4     simple    c1t1d0s4     rootdg       online
c3t1d0s2     sliced    -            -            online
c3t3d0s2     sliced    -            -            online
c3t4d0s2     sliced    -            -            online
c3t5d0s2     sliced    -            -            onlin
c3t6d0s2     sliced    -            -            online

# vxdmpadm getsubpaths dmpnodename=c3t5d0s2
NAME         STATE     PATH-TYPE  CTLR-NAME  ENCLR-TYPE   ENCLR-NAME
====================================================================
c3t5d0s2     ENABLED     -        c3         Disk         Disk
c3t2d0s2     ENABLED     -        c3         Disk         Disk
# dd if=/dev/zero of=/dev/vx/rdmp/c3t5d0s4 bs=128k count=1000
1000+0 records in
1000+0 records out
# iostat
  device    bps     sps    msps

....
  c3t5d0   4076    15.9     1.0
....
  c3t2d0   4076    15.9     1.0

Notice that I/O is going through both devices, which is incorrect, and will lead to unexpected 
data corruption.

WORK AROUND:

Just before changing (swapping) the disk slots, the multipathing should be disabled. 
That is, the library definition and/or the jbod definitions should be removed.  
The disks will then be added as single-pathed disks, and existing multi-pathing will 
be lost as a result of dmpnodes being split. Multipathing can be later re-enabled, 
and vxdctl enable will merge the correct dmpnodes.

The steps are:

(a) Exclude the library or remove the JBOD definition:
         # vxddladm excludearray libname=<> or # vxddladm rmjbod vid=<>
(b) Scan the disks for vxvm:
         # vxdisk scandisks or vxdctl enable

    (The main purpose of steps (a) & (b) is to force all the dmpnodes from the array to 
    be single-path'ed.)

(c) Include the library or JBOD definition.  
         # vxddladm includearray libname=<> or # vxddladm addjbod vid=<>
(d) Scan the disks for vxvm:
         #vxdisk scandisks or vxdctl enable.

   (The main purpose of steps (c) & (d) is to re-group the paths from scratch to again 
   form multi-path'ed dmpnodes for the luns.)

Note: The above steps will also fix the problem, if it has already occurred.

For example:
If you have physically moved a particular XP1024 disk (corresponding DDL library libvxxp256.so) 
to a different array slot, then you need to carry out the following steps:

# vxddladm excludearray libname=libvxxp256.so
# vxdctl enable  (or vxdisk scandisks)
# vxddladm includearray libname=libvxxp256.so
# vxdctl enable  (or vxdisk scandisks)

CAUTION: Once multi-pathing is disabled, the same disk is exposed as two
different disks with the potential of the disk being initialized through
the other path. Therefore, this state should not be continued for longer
than necessary. Also, the system must not be rebooted - the DG import would
then fail due to duplicate disk ids.

Solution(s)

  • sunpatch-solaris-112385

With Rapid7 live dashboards, I have a clear view of all the assets on my network, which ones can be exploited, and what I need to do in order to reduce the risk in my environment in real-time. No other tool gives us that kind of value and insight.

– Scott Cheney, Manager of Information Security, Sierra View Medical Center

;