[Gluster-devel] crashing glusterfsd consistently during benchmark / stresstesting

Daniel Maher dma+gluster at witbe.net
Fri May 9 13:58:42 UTC 2008


Hello,

I have been running some automated testing using a combination of
Bonnie++ and FFSB.  Unfortunately, i appear to be able to cause
glusterfsd to crash (and, on occasion, lock the server) on a reasonably
regular basis as a result. :(

My environment is described almost exactly by the howto i wrote for the
wiki, including server and client configs, here :
http://www.gluster.org/docs/index.php/High-availability_storage_using_server-side_AFR

My automated testing script is very simple : i simply run a series of
FFSB tests, then a Bonnie++ test, continuously on both clients.  After
two or three test cycles, one of the glusterfsd processes in the
storage cluster will crash.  A sample of the debug log from one of
these crashes is available on the pastebin :
http://glusterfs.pastebin.com/m5ecb502f

My Bonnie++ command line is as follows :
$ bonnie++ -d [mountpoint] -x 1 -q

I am using more or less the same FFSB configs available on this page :
http://tastic.brillig.org/~jwb/zfs-xfs-ext4.html

While i realise that this sort of testing is designed to be highly
academic, it is somewhat troubling that it can cause glusterfsd to
crash so thoroughly. :/

As always, i'll be happy to provide more information as requested.


-- 
Daniel Maher <dma AT witbe.net>





More information about the Gluster-devel mailing list