[LU-8463] OSSes drop into KDB during recovery

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Affects Version/s: Lustre 2.5.3
• Severity: 2
• Assignee: nasf (Inactive)
• Reporter: Herbert Yeung

Description

During the crashes reported in LU-8462, several times upon recovery, the OSS will drop into KDB. Here's a expanded timeline:

Tues Jul 26 - 394 - 2028 KDB (I/O)

395 - 2000 KDB (I/O), (disk) crash during recovery, until e2fsck
Fri jUl 29 - s393 - 0910 KDB (aborting journal)
Fri Jul 29 - s394 - 1922 KDB (disk), (disk) crash druing recovery, e2fcsk
Fri Jul 29 - s393 - 2123 KDB (disk), (disk) crash during recovery, no e2fsck
Sun Jul 31 - s393 - 1519 KDB (disk)

After the initial crash during recovery on Tuesday, I ran e2fsck on the node and was able to successfully mount the OSTs. However, running e2fsck may not be necessary as on one occassion, it was able to recovery after a second attempt.

Issue Links

• is related to: LU-8462 OSS keeps dropping into KDB