Network is down

[Love]

Harald Barth <haba at pdc.kth.se> writes:

> >> 
> >> From where are you accessing this? Nada has blocked AFS access from
> >> the outside because of a bug in AFS.
> >> 
> >
> >What bug is this? And do you know if it has been fixed in OpenAFS?
> 
> The server-runs-out-of-threads-if-routing-works-only-in-one-direction
> bug. I think this has been refered to a "server meltdown", too. Others
> who understand the problem better may fill in more facts here.
> 
> No, I don't know if it has been fixed by IBM (I suspect not because
> NADA still firewalls). I don't know if it is a problem or has been
> fixed in OpenAFS because the folks running the nada.kth.se cell don't
> want to move to OpenAFS (for a bunch of reasons) on their (solaris)
> servers. I of course argued that it can not get worse with OpenAFS ;-)

I've expericed the "meltdown" with most version of transarc afs-3.6.
Hartmut Reuter (author of mr-afs) reported that he had "solved" the problem
by stop using pthreads lwp/rx.

The symtoms I've seen is that fileserver returns VBUSY and when using
rxdebug there is lots of waiting connections ("N calls waiting for a
thread" string).

Love