Hello, I have a question about NFS clustering and data corruption. I
have built out a HA NFS solution using heartbeat on SLES 10.
Everything seems to be working well, but upon mounting the cluster on
my SLES client I received the error message:

Using NFS over UDP can cause data corruption.
Please refer to the WARNINGS section of the nfs(5) manual page.

I had not noticed this earlier as all but one of my NFS clients are
Debian.  Apparently running NFS over UDP on gigabit networks can cause
corruption due to packet reassembly and IP ID's wrapping. Since I must
run NFS over UDP for NFS fail over to work, the nfs man page presents
me with limited options. I can set the MTU of all of the machines and
the switch ports in between to a MTU of 9000, or I can set my IP
fragmentation time out to 2 seconds.

Has anyone else on this list run into issues with data corruption when
implementing HA NFS? Any suggestions on the proposed solutions from
the man page?

Thanks,
Ryan
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to