Dear all

Today i faced a strange problem in lvm (version 2) configuration on rhel 5.2. Our one of lvm volume automatically become read only. I tried to remount that volume with rw, but that failed by displying error that read only attribute is set on that volume. Restarting linux server solved the problem, but i am curious to know the reason of that problem.

thanks

Was it the LV (a Logical Volume internal to the Volume Group) or (one of) the Physical Volume(s) making up the VG that failed to mount r/w? If it was one of the PV, what cause is reported in your boot log?

From your cursory description it sounds like one of your PVs may be close to the end of its service life, and that it did not report a successful "spin up" to the kernel before the time-out expired. (The drive may have been "warmed up" when you did the re-boot.) If that's the case, you should replace the drive ASAP. (Note: If the drive is S.M.A.R.T., try a full smart scan of the PV for more information.)

Thanks

My disk is ok, sinnce other lv on that disk is working properly. My volume group consist one physical volume. Four LV are created on that volume group.
/dev/vg1/lv1
/dev/vg1/lv2
/dev/vg1/lv3
/dev/vg1/lv4

Following error was generated in /var/log/messages
Dec 24 16:29:19 abpdel2 kernel: EXT3-fs error (device dm-1): ext3_lookup: unlinked inode 12075010 in dir #4145153
Dec 24 16:29:19 abpdel2 kernel: Aborting journal on device dm-1.
Dec 24 16:29:19 abpdel2 kernel: ext3_abort called.
Dec 24 16:29:19 abpdel2 kernel: EXT3-fs error (device dm-1): ext3_journal_start
sb: Detected aborted journal
Dec 24 16:29:19 abpdel2 kernel: Remounting filesystem read-only


thanks

O.K. That identifies the problem. Since the problem was in the journal file, the reboot probably rewrote it, although the journal contents was probably lost (and probably not needed if the prior shutdown was clean).

Once a LV is activated, you can run fsck on it before you mount it. I've noticed that ext file systems created on a LV often do not have an automatic fsck run specified. Try using tune2fs to specify a -c 30 (or however many days you wish to set) so a fsck will be run when you mount. (tune2fs can also tell when your last fsck on the LV was done if you use the -l argument.)

Anyhow, an unlinked inode usually means a read failure or other fs problem that e2fsck can often resolve.

I suspect that you know all this, but others may read this thread.

Thanks for suggestion. Now i got reason of problem.