> It comes back fine. However, the amount of messages in the log for
> connection refused is much higher than in the case of ripped-out-cable
> or kswapd-won't-give-adequate-size-hunks-of-RAM-for-the-ip-fragment-
> buffer-to-grow.
Perhaps we should be setting a flag in the client structure when an
error 111 comes in, and not reporting it more than once until a
subsequent RPC call succeeds? If you had a lot of NFS work in progress
when the server rebooted you probably would see lots of complaints.
> And it was the RPC: apparent errors which worried me. These should be
> labeled as warnings or fixed.
Agreed that the messages seem rather dire -- it may be that other
strange problems will trigger the condition as well.
Regards,
Bill