[LU-12439] Convert "Response timed out..." in lnet_finalize_expired_responses to CDEBUG

Tue, 30 Jul 2019 13:38:37 +0000

I've noticed that this error message causes a lot of noise when a router node goes down. For example:

I rebooted two routers on a system with just 40 compute nodes. The first of these error messages popped up about 1 minute or so after I initiated a reboot of the routers:

Reboot started at - Fri Jun 14 14:34:35 CDT 2019

saturn-smw:/var/opt/cray/log/p2-current # grep -m 1 lnet_finalize_expired_responses console-20190614
2019-06-14T14:35:33.704182-05:00 c0-1c1s9n3 LNet: 10316:0:(lib-move.c:2888:lnet_finalize_expired_responses()) Response timed out: md = ffff8810119a32a8: nid = 485@gni4

In the time it took the routers to reboot, about 8 minutes, there were 797 entries from lnet_finalize_expired_responses in the console log:

saturn-smw:/var/opt/cray/log/p2-current # grep -c lnet_finalize_expired_responses console-20190614
797
saturn-smw:/var/opt/cray/log/p2-current #

I don't see much value from this message for system administrators, so I think it should be converted to a CDEBUG

Whamcloud Community JIRA

[LU-12439] Convert "Response timed out..." in lnet_finalize_expired_responses to CDEBUG