Peter Kratzer <[EMAIL PROTECTED]> writes: > We have also contacted the StorageTek support. They don't have a clue > either. They suggested a network trace on the storage device. The problem > there is the trace stopa a file size of 1 GB. And that's nothing for lots > of users using the same machines and often producing hundreds of gigabytes > of hardware simulation data. As long as I cannot predicably reproduce the > problem I am lost there.
Going back and reading the original symptoms, I am suspicious that somehow the NFS idempotency cache is getting in your way. This is what comes into play when a packet is lost (are you running NFS or UDP or TCP, by the way?) and the client retries. However, this is just a suspicion. In any event, I doubt whether it's a bug in coreutils; more likely it's a bug in the NFS client code on the Linux side, or in the NFS server code on the StorageTek side. If you have some non-Linux clients (e.g., FreeBSD, or the reference platform Solaris) you can test part of this theory by running the same version of coreutils on both and seeing whether the problem occurs on them. _______________________________________________ Bug-coreutils mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-coreutils
