> On Jun 19, 2024, at 3:32 AM, Harald Dunkel <harald.dunkel@xxxxxxxxxx> wrote: > > On 2024-06-18 16:52:00, Chuck Lever wrote: >> I can find no NFSD or SUNRPC changes between v6.1.85 and v6.1.90. >> Between v6.1.76 and v6.1.85, there are 48 NFSD changes, 8 changes to > > fs/nfs/, and 3 changes to net/sunrpc/. > > > Are there any NFS client changes during the second half of May that >> correlate with the NFSD misbehavior? > There was a big change: Most clients (>100) were upgraded from Debian > 11 to Debian 12 within the last couple of months. A big chunk of > about 40 hosts have been upgraded on May 21st. This included the > kernel upgrade to 6.1.xx, and the upgrade of the "nfs-common" package > from version 1.3.4 to 2.6.2. > > The current nfs-common package provides these tools: > > /sbin/mount.nfs > /sbin/mount.nfs4 > /sbin/rpc.statd > /sbin/showmount > /sbin/sm-notify > /sbin/umount.nfs > /sbin/umount.nfs4 > /usr/sbin/blkmapd > /usr/sbin/mountstats > /usr/sbin/nfsconf > /usr/sbin/nfsidmap > /usr/sbin/nfsiostat > /usr/sbin/nfsstat > /usr/sbin/rpc.gssd > /usr/sbin/rpc.idmapd > /usr/sbin/rpc.svcgssd > /usr/sbin/rpcctl > /usr/sbin/rpcdebug > /usr/sbin/start-statd > > I have 3 clients with more recent kernel versions 5.10.209, 5.10.218 > and 6.7.12. The 5.10.x hosts are running Debian 12 and this kernel > for several months, but the host with 6.7.12 is running this kernel > since May 16th. This makes it very difficult to say what might have introduced the problem in your environment. My next step might be capturing network traces, but you said the problem is hard to reproduce and you seem to have a lot of clients, making such capture impractical. > These "backports" kernels were installed for testing purposes. Would > you recommend to downgrade these kernels to 6.1.90? > > As asked before, do you think the problem could be related to running > nfsd inside an LXC container? There's no way to answer either of these questions since we have no idea what the problem is yet. -- Chuck Lever
