[LU-1065] High rate of obd_ping failure with client <-> OST evictions

• Type: Bug
• Resolution: Duplicate
• Priority: Major
• Affects Version/s: Lustre 1.8.x (1.8.0 - 1.8.5)
• Environment: Clients: Lustre Version: 1.8.6.81; Servers: 2.1.0 with ofed 1.5.3.1
• Severity: 3
• Assignee: Hongchao Zhang
• Reporter: Mahmoud Hanafi

Description

We are seeing a large amount of obd_ping failures then client eviction from our 2.1 servers relative to our 1.8.6/5 servers.
We have one 2.1 filesystem and six 1.8.x filesystems.
Here you can see obd_ping counts for each data:
---- lustre-20120125 -----
nbp5 1022
allothers 24
---- lustre-20120126 -----
nbp5 760
allothers 33
---- lustre-20120127 -----
nbp5 420
allothers 6
---- lustre-20120128 -----
nbp5 226
allothers 97
---- lustre-20120129 -----
nbp5 36
allothers 7
---- lustre-20120130 -----
nbp5 243
allothers 19
---- lustre-20120131 -----
nbp5 808
allothers 17
---- lustre-20120201 -----
nbp5 81
allothers 2

Attached are typical client log(r8610n14.lustre)
and server logs(service151.lustre.feb1)

Issue Links

• duplicates: LU-874 Client eviction on lock callback timeout
• Trackbacks: Lustre 1.8.x known issues tracker While testing against Lustre b18 branch, we would hit known bugs which were already reported in Lustre Bugzilla https://bugzilla.lustre.org/. In order to move away from relying on Bugzilla, we would create a JIRA

Comments

[Peter Jones]

Assumed to be a duplicate of LU-874. We will reopen if this proves to not be the case

[Hongchao Zhang]

Hi Mahmoud

have you tested it with the patch in LU-874? what is the result?
Thanks

[Peter Jones]

Hi Hongchao

Could you please look into this one?

Thanks

Peter

[Andreas Dilger]

This issue is already being tracked under LU-874, which has a number of patches scheduled to land for the 2.1.1 release.