[OpenAFS] "These servers unavailable due to network or server problems"

[Logan O'Sullivan Bruns]

I saw a similar behavior a few days ago. One client complained that
the server was down but from every other machine the server looked
fine. In my case it appears to be related to a resource leak in the
1.6.1pre2 client and restarting with a clean cache fixed it.

I'd be curious to see if you are running into the same issue. If so I
think you might see your cache growing excessively. If easy you might
try:

sudo du -s -k /var/db/openafs/cache
fs getcacheparms

In my case it seems to either behave like you saw or fill up the cache
entirely and panic. (Mine is using a dedicated ufs partition though.)
Here, for example, even after increasing my cache partition to double
its original size it still eventually fills up.

$ df -k /usr/vice/cache
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/zvol/dsk/rpool/afscache
                       4130285   4085457      3526 100% /usr/vice/cache
$ fs getcacheparms
AFS using 1357687 of the cache's available 1800000 1K byte blocks.

  - logan

On Thu, Feb 09, 2012 at 10:56:30AM +0100, Rainer Toebbicke wrote:
> Hello,
> 
> I'm having trouble with 1.6.1pre2 on Mac OS 10.6.8:
> 
> Symptom:
> 
> "ls /afs/cern.ch/project/swift" says "swift: Operation timed out"
> "fs checkservers" says "These servers unavailable due to network or server problems:  afs43.cern.ch"
> "cmdebug localhost" doesn't say anything
> 
> "fs checkservers; fs checkvolumes" does not help, nor does "fs flush" on the path or it's parent
> 
> What is surprising is that AFS is so sure that the server is down: nobody else sees it down, for the past 30 minutes AFS didn't send a single packet to it despite the "checkservers" in a loop. 
> 
> Any idea? Other than cycling AFS?
> 
> Cheers, Rainer_______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-info