Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
The filesystem was created correctly on the drbd0 device and verified that this is the case. This is the same issue as shown in: http://lists.linbit.com/pipermail/drbd-user/2005-October/003860.html This allusive bug has never gone away (I've been experiencing this for almost two years now). I would like to know how people are dealing w/ this. It has recently become a real problem in my production environment (almost every day). I created the filesystem on both production setups on the DRBD device. They might seem to fail (but not always) when heavily used (lots of web activity or during backups). Brent At 08:37 AM 7/10/2006 +0200, Luc de Louw wrote: >Brent Jensen wrote: >>Has anyone learned anything more about the drbd file corruption issues? >>It has started to become a big problem now. I have two different servers, >>different motherboards, both running scsi disks on an md0 device. I'm >>getting them both to go read-only. About every couple of days (and >>sometime multiple times a day) I'm getting this error. >>I'm using drbd version 0.7.17 on both. Might there be a fix in a later >>release? >>Server 1 >>======== >>Centos 4.3 (2.6.9-34.0.1.ELsmp) >>Jul 8 11:16:34 nfs5 kernel: EXT3-fs error (device drbd0): ext3_readdir: >>bad entry in directory #9339548: rec_len is smaller than minimal - of >>fset=512, inode=3614863868, rec_len=0, name_len=0 >>Jul 8 11:16:34 nfs5 kernel: Aborting journal on device drbd0. >>Jul 8 11:16:34 nfs5 kernel: ext3_abort called. >>Jul 8 11:16:34 nfs5 kernel: EXT3-fs error (device drbd0): >>ext3_journal_start_sb: Detected aborted journal >>Jul 8 11:16:34 nfs5 kernel: Remounting filesystem read-only >[..] > >Did you created the filesystem on /dev/md0 or on /dev/drbd0? Looks like >the filesystem was created on /dev/md0 which is wrong because the metadata >(if using metdata-internal) needs additional 128MB space. > >rgds > >Luc