Subject: Re: NFS timeouts (sorta weird)
From: Peter Seebach <email@example.com>
Date: 12/17/2002 22:26:16
In message <200212180000.gBI00Tk23476@guild.plethora.net>, Peter Seebach writes
>In message <20021217233141.GD19317@deathmitten.example.org>, John Franklin wri
>>Unless you've changed it in your kernel config file (options BUFCACHE),
>>then it is 5% of total memory. Your NFS server may be trying to get
>>enough bufs to stat the inodes and create the NFS request cache.
>Exceptionally likely-sounding! I will look at this.
Nope, it's just mundane packet loss. For some reason, about half of all
mount attempts just get lost. If I run tcpdump on both ends, when the
mount fails, the client sends two packets right away, and two more a while
later, and none of them get through to the other side. It doesn't appear
to ever recover - but if I restart it, it sometimes works.
No packets other than NFS mount and unmount requests are staying lost.
It appears that "mount_nfs -T" still implies udp for the mount protocol
itself. I don't suppose there's a way to override this?