[OpenAFS-devel] volserver dies on linux

[Kevin Coffman]

What was the original failure?

I've seem problems in the past where the LWP on the volume server is 
looping through the vnode index file deleting files and 'forgets' to 
let the original LWP run to tell the client it is still alive and 
processing the request.  But this usually gets straightened out when 
the volume server finally finishes the request.  (Never seen the 
SIGABRT.)

> 
> Are folks doing development aware of a volserver problem that manifests itself 
> when trying to delete several volumes in succession?  The scenario looks 
> something like:
> 
> - Create a couple thousand volumes on a 300+ GB (ext3 on software raid 5) 
> partition
> 
> - Put some Large Files (i.e. 500 MB to 1 GB) in some of the volumes, but leave 
> most empty
> 
> - Start deleting the volumes
> 
> After a few dozen, the volserver starts rejecting the 'vos remove' commands, 
> although the volumes disappear from the vldb.  When I tried to go back over 
> the data with 'vos zap' the error message was:
> 
> Could not start transaction on volume 2004066564
> Volume not attached, does not exist, or not on line
> Error in vos zap command.
> Volume not attached, does not exist, or not on line
> 
> I get that message for as many more 'vos zap' commands as I try.  After that 
> point, if I 'vos listvol' the partition, the volserver dies with SIGABRT.  
> When it restarts, I can delete a few dozen more volumes before the volserver 
> goes bad again.
> 
> Needless to say, this isn't a very efficient way to delete thousands of 
> volumes.
> 
> This is OpenAFS 1.2.2 on Red Hat 7.2 with kernel.org 2.4.18 kernel, dual 
> Athlon file server.
> 
> Has this already been diagnosed, or should I investigate further?
> 
> --
> Dave Thompson  <thomas@cs.wisc.edu>
> 
> Associate Researcher                    Department of Computer Science
> University of Wisconsin-Madison         http://www.cs.wisc.edu/~thomas
> 1210 West Dayton Street                 Phone:    (608)-262-1017
> Madison, WI 53706-1685                  Fax:      (608)-262-6626
> --
> 
> 
> 
> 
> _______________________________________________
> OpenAFS-devel mailing list
> OpenAFS-devel@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-devel
>