#46 fsck prompt during failover

closed-fixed
nobody
Filesystem (49)
3
2005-08-26
2004-04-14
No

I was testing a had a failover which caused the fsck to
stop and prompt:

<4>Node 3 has gone down!!!
INIT: Sending processes the TERM signal
fsck 1.32 (09-Nov-2002)
e2fsck 1.32 (09-Nov-2002)
ext3 recovery flag clear, but journal has data.
Run journal anyway<y>?

1. Clearly, in simple cases we should not prompt the user.
2. If there is a severe problem, we should abort the
fsck and NOT remount the filesystem until
administrative intervention occurs.
3. Is this particular problem related to the orphan
inodes and an improper umount?

Discussion

  • David Zafman

    David Zafman - 2004-05-10
    • assigned_to: dzafman --> nobody
     
  • David Zafman

    David Zafman - 2004-05-10

    Logged In: YES
    user_id=297844

    Maybe we need to make the nodedown version of fsck work like
    the root fsck (returns an error instead of prompting). We
    also need to figure out what to do if the fsck needs manual
    intervention. How do we let the admin. handle the problem
    later?

     
  • Bruce J. Walker

    Bruce J. Walker - 2004-08-25

    Logged In: YES
    user_id=296932

    Can we invoke fsck -y (for yes) and would that help?

     
  • Roger Tsang

    Roger Tsang - 2005-08-26
    • status: open --> closed-fixed
     
  • Roger Tsang

    Roger Tsang - 2005-08-26

    Logged In: YES
    user_id=1246761

    Latest ckroot.ssi does /sbin/fsck -y -t $ROOTFS $ROOTDEV and
    will abort on severe error. Admin will have to do recovery
    from console on severe errors.

     
  • Kari Hurtta

    Kari Hurtta - 2005-08-27

    Logged In: YES
    user_id=495208

    ( Side note: I think that some systems use fsck -p
    on boot. -p option is on e2fsck as:
    Automatically repair ("preen") the file system without
    any questions.
    )

     

Log in to post a comment.