[LU-11383] A lot of cfs_fail_loc error message for Fake I/O

• Status: Resolved
• Project: Lustre
• Type: Improvement
• Priority: Minor
• Reporter: Li Xi
• Assignee: Li Xi
• Resolution: Duplicate

Description

Usually, enabling fake I/O means there will be benchmark running on the system. However, a lot of messages are printed to the console:

 

Sep 17 05:50:49 es200nv2-vm1 kernel: systemd-journald[690]: /dev/kmsg buffer overrun, some messages lost.
Sep 17 05:50:49 es200nv2-vm1 kernel: Lustre: *** cfs_fail_loc=238, val=0***
Sep 17 05:50:49 es200nv2-vm1 kernel: Lustre: *** cfs_fail_loc=238, val=0***
Sep 17 05:50:49 es200nv2-vm1 kernel: Lustre: *** cfs_fail_loc=238, val=0***
Sep 17 05:50:49 es200nv2-vm1 kernel: Lustre: *** cfs_fail_loc=238, val=0***
Sep 17 05:50:49 es200nv2-vm1 kernel: Lustre: *** cfs_fail_loc=238, val=0***
Sep 17 05:50:49 es200nv2-vm1 kernel: Lustre: *** cfs_fail_loc=238, val=0***

 

This is annoying, because a lot of CPU resource will be used by message printing.

Issue Links

• duplicates: LU-11373 CERROR/CWARN messages are not throttled
• is duplicated by: LU-11384 wrong check in libcfs_debug_vmsg2()

Comments

[Gerrit Updater]

Li Xi ([e-mail]) uploaded a new patch: https://review.whamcloud.com/33174
Subject: LU-11383 ofd: keep fake IO quiet
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 96453ee4682047196ecce4aef4db379bb9ad73f2

[Andreas Dilger]

There is something broken with the CDEBUG rate limiting in libcfs in master. It shouldn't print so many messages.

[Li Xi]

Yeah, that makes sense. I didn't notice so many messages until recently.