[LU-17610] Batched statahead errors on client against 2.12 server

• Type: Bug
• Resolution: Unresolved
• Priority: Minor
• Affects Version/s: Lustre 2.16.0
• Environment: Client: ; - CentOS 8.0.1905 ; - Kernel 4.18.0-80.11.2.el8_0.x86_64 ; - Lustre 2.15.61 ; ; Server: ; - CentOS 7.9.2009 ; - Lustre 2.12.6
• Severity: 3
• Assignee: WC Triage
• Reporter: Mark Dixon

Description

Mounting a 2.12 lustre filesystem on a 2.15.61 client generates a message of the following form when running "/bin/ls -F" in any uncached directory with regular files in it:

Mar  5 12:11:25 xxxxx kernel: LustreError: 11-0: lustre-MDT0000-mdc-ffff9c53d793a800: operation mds_batch to node 172.18.16.22@o2ib failed: rc = -524
Mar  5 12:11:25 xxxxx kernel: LustreError: 4642:0:(layout.c:2309:__req_capsule_get()) @@@ Wrong buffer for field 'batch_update_reply' (1 of 1) in format 'MDS_BATCH', 0 vs. 8 (server)  req@0000000021c4d84b x1792688163884672/t0(0) o63->lustre-MDT0000-mdc-ffff9c53d793a800@172.18.16.22@o2ib:24/4 lens 3440/192 e 0 to 0 dl 1709640701 ref 1 fl Interpret:RQU/200/0 rc -524/-524 job:'' uid:0 gid:0
 

git bisect says that this started appearing due to LU-15562, which among other things enabled statahead by default.

I don't know if the message just means that the client is talking to a non-statahead capable server, in which case it would be nice if the client log does not fill up with these messages please : )

Issue Links

• is related to: LU-17783 Wrong buffer for field 'batch_update_reply' in format 'MDS_BATCH' from statahead for old server
• is related to: LU-15562 statahead: using try lock for batched statahead to avoid deadlock with rename and migrate operations