On Sunday 29 November 2009 21:33:35 Tom Feiner wrote: > Thanks for the detailed bug report! And sorry for taking so long to answer > the bug, I've been working hard lately getting munin 1.4 package ready :).
Thanks answering :) > Have you discovered anything else since you've reported the bug? Are other > processes suffering from the same problem as munin-graph? The reason I'm > asking is that AFAIK, processes that run into disk sleep state and stay > there, are either: > > * Trying to communicate with a failed NFS/some other remote filesystem. > * Trying to communicate with a failed/failing disk drive. > * Trying to write to a filesystem mounted using a new/non-stable driver? > > Are any of the filesystems that munin-graph writes to NFS based? Or some > other remote filesystems / external drives / drives mounted with a new > driver which might have problems? Everything's ext3, internal hard disk. However I've got 4 swap files /home/swapfile1 swap swap defaults 0 0 /home/swapfile2 swap swap defaults 0 0 /home/swapfile3 swap swap defaults 0 0 /home/swapfile4 swap swap defaults 0 0 Maybe that the problem came from that, since in the call stack there is "system_call_after_swapgs", but I'm just guessing. Also, I have a pretty low confidence in the quality of my hard disk, since it's a cheap dedicated host, the hardware in the server is probably of very poor quality. It might be a hardware failure and there is just nothing to be done... (I know it's useless to have 4 distinct swap files, but I actual had a problem with svn that turned out to use an increasing amount of RAM in a very short time, so I added some swap on the fly to avoid an OOM, but I did not remove them afterward). > Is this problem reproducible? And if so, can you attach an strace of the > process? No I'm sorry I could not figure out how this happened, this had just appeared by itself in normal operation, I did not touch anything in any configuration of the server for days... -- Rémy Sanchez http://hyperthese.net
signature.asc
Description: This is a digitally signed message part.