Server crashes only when quota is "on"

  • Hello,

    I have a SuSE 9.3 Server with dquot_6.5.1 at a RAID 5 via software on a ReiserFS filesystem /home (206GB) dir using 4 SCSI drives. I have about 200 linux(nfs)/windows clients. The server is an Dual Xeon with 5 GB RAM. It Runs Apache/Samba/NFS/LDAP/Postfix.

    Sometimes (It works for 2~3 days without any problem), the number of processes of Samba (samba-3.0.13-1.1) starts to grow and the server stops do answer. So, I have to reboot it (At boot time, the partition can't be umounted and I have to turn off the power). When it is working, it is always fast. The problem occurs suddenly.
    I observed that it only happens when diskquota is "on". If I don't turn on the quota, the server works without any problem.

    Do you have any hint or suggestion of test that I could do to solve this problem?

    Best Regards,

    • Jan Kara
      Jan Kara

      Hmm, I guess the problem is with deadlocks in quota code in kernel (I guess you'll see Samba processes hanging in D state). The only real fix is to upgrade kernel to something newer which has the deadlocks fixed. I don't recall when the fix went in exactly but I think it is already in 2.6.9 kernel.