DMP enhancement to handle failing IO when it is not able to interpret sense data( without fix possible side effects like dgdisabled diskgroups, disk failures)
Article ID: 100023280
Updated On:
Resolution
DMP is very aggressive when failing IO when it is not able to interpret the sense data.
Without the below enhancement, diskgroup's may go into a "DGDISABLED" state due to what appears to be a SAN outage.
Sample messages
/etc/vx/dmpevents.log
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d39s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d38s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d34s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d14s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d116s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
The issue is that the SCSI inquiry status encountered cdb status STATUS_MET (0x04) which is not currently handed without the enhancement, hence the error analysis treats
the error as a "generic device error".
DMP_NODE_ERR_EVENT
Media error occured on Dmpnode
This event is raised by DMP when it detects a media error, i.e. the connectivity to the LUN is fine but the I/O operation could not be completed.
The administrator needs to check the status of the LUN on the array.
Other messages
Tue Aug 25 21:43:20.793: I/O error occured on Path c3t5006048AD52DC5C6d114s2 belonging to Dmpnode c3t5006048AD52DC5C6d114s2
Tue Aug 25 21:43:20.799: Disabled Path c3t5006048AD52DC5C6d114s2 belonging to Dmpnode c3t5006048AD52DC5C6d114s2
Tue Aug 25 21:43:20.803: Disabled Dmpnode c3t5006048AD52DC5C6d114s2
Tue Aug 25 21:43:20.805: I/O analysis done on Path c3t5006048AD52DC5C6d114s2 belonging to Dmpnode c3t5006048AD52DC5C6d114s2
Tue Aug 25 21:43:20.816: I/O error occured on Path c3t5006048AD52DC5C6d111s2 belonging to Dmpnode c3t5006048AD52DC5C6d111s2
Note: Unfortunately the code was too complex to backport to 5.0 MP1 due to numerous potential code regression possibilities.
Patch Information:
PRODUCT VERSION PLATFORM TYPE PATCH NAME
----------------------------------------------------------------------------------------------------------------------------------------------------------------
Veritas Volume Manager 5.0MP3 Solaris x86 RP vm-sol_x64-5.0MP3RP2
Veritas Volume Manager 5.0MP3 Solaris RP vm-sol_sparc-5.0MP3RP2
Veritas Volume Manager 4.1MP2 Solaris RP vm-sol_sparc-4.1MP2RP4
Veritas Volume Manager 4.1MP2 Solaris HF vm-sol_sparc-4.1MP2RP3HF17
Veritas Volume Manager 4.1MP2 Solaris HF vm-sol_sparc-4.1MP2RP3HF16
Without the below enhancement, diskgroup's may go into a "DGDISABLED" state due to what appears to be a SAN outage.
Sample messages
/etc/vx/dmpevents.log
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d39s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d38s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d34s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d14s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
Tue Aug 25 19:26:07.449: SCSI error occured on Path c7t5006048AD52DC5C9d116s2: opcode=0x12 reported scsi transport error (status=0x4, key=0x0, asc=0x0, ascq=0x0)
The issue is that the SCSI inquiry status encountered cdb status STATUS_MET (0x04) which is not currently handed without the enhancement, hence the error analysis treats
the error as a "generic device error".
DMP_NODE_ERR_EVENT
Media error occured on Dmpnode
This event is raised by DMP when it detects a media error, i.e. the connectivity to the LUN is fine but the I/O operation could not be completed.
The administrator needs to check the status of the LUN on the array.
Other messages
Tue Aug 25 21:43:20.793: I/O error occured on Path c3t5006048AD52DC5C6d114s2 belonging to Dmpnode c3t5006048AD52DC5C6d114s2
Tue Aug 25 21:43:20.799: Disabled Path c3t5006048AD52DC5C6d114s2 belonging to Dmpnode c3t5006048AD52DC5C6d114s2
Tue Aug 25 21:43:20.803: Disabled Dmpnode c3t5006048AD52DC5C6d114s2
Tue Aug 25 21:43:20.805: I/O analysis done on Path c3t5006048AD52DC5C6d114s2 belonging to Dmpnode c3t5006048AD52DC5C6d114s2
Tue Aug 25 21:43:20.816: I/O error occured on Path c3t5006048AD52DC5C6d111s2 belonging to Dmpnode c3t5006048AD52DC5C6d111s2
Note: Unfortunately the code was too complex to backport to 5.0 MP1 due to numerous potential code regression possibilities.
Patch Information:
PRODUCT VERSION PLATFORM TYPE PATCH NAME
----------------------------------------------------------------------------------------------------------------------------------------------------------------
Veritas Volume Manager 5.0MP3 Solaris x86 RP vm-sol_x64-5.0MP3RP2
Veritas Volume Manager 5.0MP3 Solaris RP vm-sol_sparc-5.0MP3RP2
Veritas Volume Manager 4.1MP2 Solaris RP vm-sol_sparc-4.1MP2RP4
Veritas Volume Manager 4.1MP2 Solaris HF vm-sol_sparc-4.1MP2RP3HF17
Veritas Volume Manager 4.1MP2 Solaris HF vm-sol_sparc-4.1MP2RP3HF16
LINUX:
On Linux, the DMP enhancement is addressed in a hot-fix for all the 5.0 MP3 RP branches.
5.0 RU3 also includes the update, although 5.0 MP4 is recommended for Linux.
AIX:
The DMP enhancement is included in 50MP3RP1 onwards, however, 5.0 MP3 RP4 and above is recommended.
Issue/Introduction
DMP enhancement to handle failing IO when it is not able to interpret sense data (without the patch fix possible side effects include dgdisabled diskgroups, disk failures...etc.)
Was this article helpful?
Yes
No
Powered by
