Re: NFS v3 - FreeBSD 12 server connection getting timed out on CentOS 8 client
Date: Thu, 14 Oct 2021 00:45:58 UTC
On Wed, Aug 25, 2021 at 3:52 AM parv/freebsd wrote: > On Tue, Aug 24, 2021 at 2:20 PM parv/freebsd wrote: > > Hi me, > > On Sun, Aug 22, 2021 at 8:05 PM parv/freebsd wrote: >> >> (I have rearranged my quoted email) >> >> [CentOS 8 NFS v3 client sometimes prints ...] >> >>> Aug 21 08:31:04 [CLIENT] kernel: watchdog: BUG: soft lockup - CPU#0 >>> stuck for 22s! [kworker/u24:0:672075] >>> >>> Aug 21 08:31:04 [CLIENT] kernel: watchdog: BUG: soft lockup - CPU#1 stuck for 22s! [kworker/u24:10:672632] >>> >>> ... > >> How do I go about investigating or solve this? >> >> ... > >> I have collected tcpdump output on both server & client ... >> >> https://github.com/parv/centos8-nfs3-issue/ >> > ... > > While trying to find if there is an update for CentOS 8 "nfs-utils" > package, I came upon the thread on "linux-nfs" mailing list ... > > CPU lockup in or near new filecache code, 201912, > https://marc.info/?l=linux-nfs&m=157599524312895&w=2 ; > > a patch, 201912, > https://marc.info/?l=linux-nfs&m=157609449120027&w=2 ; > > more to come, 202001 > https://marc.info/?l=linux-nfs&m=157807452103938&w=2 ? > > > Now the question is will the fix be available on CentOS 8 anytime > soon, as above thread was concerned with, AIUI, kernel 5.5 (CentOS 8 > kernel is hovering around 4.18)? > > Not a FreeBSD issue anymore, however, seems like. > ... not quite. (In addition or stead of "uname -a", I should have mentioned "freebsd-version -kur" in hindsight.) Moving to FreeBSD 12.2-p9 or -p10 did not solve issue of FreeBSD NFS v3 server timing out on CentOS 8 client after ~13-14 minutes of file copying. But going back to -p8 did; there were no more "CPU stuck", stack traces on CentOS 8. Update of -p9 brought changes to libcasper. - parv