Re: [Jfs-discussion] `diFree: numfree > numinos` error after LVM resize

[Ciprian D. C. <cip...@gm...>]

On Sun, Nov 22, 2009 at 9:48 PM, Ciprian Dorin, Craciun
<cip...@gm...> wrote:
>    Hello all!
>
>    Today I've resized my JFS partition from 50 GB to 62 GB and then
> again to 96 GB.
>
>    Each of these resizes were followed by a `mount /mnt/something -o
> remount,resize`.
>
>    But after a few minutes after the second resize, the error
> `diFree: numfree > numinos` popped out and the disk went in RO...
>
>    Any ideas? I'm mainly interested in:
>    * can I assume that right now the file system is ok? (so I can do
> an emergency backup;)
>    * should I worry that after I unmount, fskc and remount I'll loose my data?
>
>    Thanks,
>    Ciprian.

    So, it seems everything is Ok. (But almost gave me an heart attack... :) )

    So for those interested here is the outcome:
    * after the `diFree: numfree > numinos` error, the disk went read-only;
    * (as a side-note it went read-only only for the newly opened
files, because for files which were opened read-write before the error
happened (and which have not been closed after), process could still
write to them; (for example Firefox database, etc.))
    * before fsck-ing the disk, I've backed up whatever I had on the disk;
    * I've also done an md5sum on every file just to be sure;
    * unmounted the file system;
    * runned jfs_fsck in `-n` (read-only) mode to see what would
happen (a lot of errors were reported about inconsistent states for
free space mappings???);
    * runned jfs_fsck in `-a` (prime / auto correct) mode and seen the
same errors as in `-n` mode; (it took a while;)
    * mounted the file system;
    * checked the md5's and everything was Ok; (except those files
that I was mentioning they remained in RW mode...)

    Until now nothing was lost... :)

    But I've learned some lessons:
    * before messing-around with LVM or JFS (or any other file system)
on live data, do a backup; :)
    * it's better to unmount / fsck / mount -o resize / unmount / fsck
/ mount again (just to be on the safe side);
    * compare the data with what you have in the backup... :)
    (It might sound tedious, but if you care about the data it's
nothing compared with the possibility of losing it...) :)

    Ciprian.

    P.S.: If anyone is interested in the jfs_fsck report, I could send it.