[LUDOC-332] Ordering of mgsnode options critical for first-time mount

• Type: Bug
• Resolution: Unresolved
• Priority: Blocker
• Severity: 3
• Assignee: Lustre Manual Triage
• Reporter: Cliff White (Inactive)

Description

When doing mkfs.lustre on a new device, if there is a failover pair of MGS nodes, the primary --mgsnode must be specified first in the mkfs.lustre options. Otherwise, the first-time mount will fail. This needs to be clarified in the manual.
See https://jira.hpdd.intel.com/browse/AH-4 for the messy details.
If the primary node is 10.102.133.18 and the secondary is 10.102.133.19

mkfs.lustre --mdt --fsname=IRN1_C --servicenode=10.102.133.19@tcp --servicenode=10.102.133.18@tcp --mgsnode=10.102.133.19@tcp --mgsnode=10.102.133.18@tcp --index=0 /dev/mapper/mpathb
"

Will fail on first time mount
This:

mkfs.lustre --mdt --fsname=IRN1_C --servicenode=10.102.133.19@tcp --servicenode=10.102.133.18@tcp --mgsnode=10.102.133.18@tcp --mgsnode=10.102.133.19@tcp --index=0 /dev/mapper/mpathb
"

will work.

Comments

[Richard Henwood (Inactive)]

NOTE: MGS failover is not addressed in the manual currently

https://build.hpdd.intel.com/job/lustre-manual/lastSuccessfulBuild/artifact/lustre_manual.xhtml#understandingfailover