Volume Manager does not recover a failed path on 5.1SP1RP2 for Solaris

Article: 100007483
Last Published: 2012-12-07
Ratings: 0 0
Product(s): InfoScale & Storage Foundation

Problem

 In the Solaris version of 5.1SP1RP2 DMP (Dynamic Multi Pathing), the driver does not automatically enable the failed paths of Logical Units (LUNs) that are restored after some maintenance activities such a Fibre Channel switchport disable event or cable removal from the HBA.

 After installing RP2 for 5.1SP1 (Solaris Sparc and x86 only), the restore deamon fails to automatically re-enable failed (disabled) paths and the paths remain in disabled state until rescanned with 'vxdctl enable' or 'vxdisk scandisks' command.

Error Message

 

1) DMP path restoration set up.apg_sol03 / # vxdmpadm start restoreapg_sol03 / # vxdmpadm stat restoredThe number of daemons running : 1The interval of daemon: 300The policy of daemon: check_disabledapg_sol03 / # vxdmpadm getsubpaths dmpnodename=ams_wms0_17NAME STATE[A] PATH-TYPE[M] CTLR-NAME ENCLR-TYPE ENCLR-NAME ATTRS================================================================================c2t50060E8010429A91d17s2 ENABLED SECONDARY c2 AMS_WMS ams_wms0 -c3t50060E8010429A90d17s2 ENABLED(A) PRIMARY c3 AMS_WMS ams_wms0 -apg_sol03 / #2) The HBA path was failed (cable pull or switch port disable) in this example at "Nov 10 15:12:28"##### snip /var/adm/messages #####Nov 10 15:12:28 apg_sol03 emlxs: [ID 349649 kern.info] [ 5.0301]emlxs0: NOTICE: 710: Link down.##### snip /var/adm/messages #####apg_sol03 / # vxdmpadm getsubpaths dmpnodename=ams_wms0_17NAME STATE[A] PATH-TYPE[M] CTLR-NAME ENCLR-TYPE ENCLR-NAME ATTRS================================================================================c2t50060E8010429A91d17s2 DISABLED SECONDARY c2 AMS_WMS ams_wms0 -c3t50060E8010429A90d17s2 ENABLED(A) PRIMARY c3 AMS_WMS ams_wms0 -apg_sol03 / #3) The HBA path was restored at "Nov 10 15:16:34".##### snip /var/adm/messages #####Nov 10 15:16:34 apg_sol03 emlxs:[ID 349649 kern.info][5.0505]emlxs0:NOTICE: 720: Link up.(2Gb, fabric, initiator)##### snip /var/adm/messages ##### DMP path status does not return to "ENABLED" after the expected restore cycle time has elapsed(default: 300 Seconds) after reconnection to the fabric.apg_sol03 / # date;vxdmpadm getsubpaths dmpnodename=ams_wms0_17Thu Nov 10 16:47:36 JST 2011NAME STATE[A] PATH-TYPE[M] CTLR-NAME ENCLR-TYPE ENCLR-NAME ATTRS================================================================================c2t50060E8010429A91d17s2 DISABLED SECONDARY c2 AMS_WMS ams_wms0 -c3t50060E8010429A90d17s2 ENABLED(A) PRIMARY c3 AMS_WMS ams_wms0 -apg_sol03 / #

 

 

 

Cause

 A regression is introduced via dmp_check_open() which fails every time restore deamon is kicked. As a result paths are not getting re-enabled. The open() was issued on a wrong device partition which resulted in failed status at DMP layer though it was enabled in the array side.

It is applicable to all array types.

 

Solution

Engineering has fixed this issue in Volume Manager Hotfix 5.1SP1RP2P2HF3  for both Solaris Sparc and x86 platforms.

If your version is currently 5.1SP1RP2, you must apply 5.1SP1RP2P2 patch as a prerequisite to applying the Hotfix.

Contact Veritas Technical Support to obtain the Hotfix.

As a workaround issuing 'vxdisk scandisks' or 'vxdctl enable' will re-enable failed paths.

 

Applies To

VxVM 5.1SP1RP2 on Solaris 10 SPARC and x86

References

Etrack : 2635476

Was this content helpful?