UNSOLVED

jjgrinwis

updated

15 years ago

J

jjgrinwis

48 Posts

0

1198

May 5th, 2011 01:00

fs_dedupe stuck at 95% ?

Hi,

we're deduping a 2.5TB fs and it's been busy for a couple of days.

Since yesterday, it looks like it's stuck at 95%, Scanning status is not being updated anymore.

server_df of of this FS is still showing a decrease in space used and as it's the only process makeing use of this FS, it can only be the dedupe process compressing/dedupping the data. When we also check the server_stats, we're also seeing an average of around 6MB/s read and write to this FS.

Anyone any idea why it's not updating the Scanning status.

Regards,

John

  • dynamox

    11 Legend

    20419 Posts

    87439 Points

    451

    0

    Posted May 5th, 2011 06:00

    what if you suspend dedupe and then resume it ?

  • jjgrinwis

    48 Posts

    451

    0

    Posted May 5th, 2011 07:00

    Just checked and dedupe finished this afternoon.

    $ fs_dedupe -l
    id      name       state     status                        time_of_last_scan                     original_data_size  usage space_saved
    172    FS          On        Idle                            Sat Apr 30 15:34:31 CEST 2011 2428172 MB         86%    730231 MB (30%)
    It took the datamover couple of days but customer has saved 30% on this FS.
    Looks like dedupe process has some issues updating the stats for the last 5%,
    Regards,
    John
  • Rainer_EMC

    6 Operator

    8645 Posts

    451

    0

    Posted May 5th, 2011 07:00

    Did you look at the data mover logs via server_log ?

  • jjgrinwis

    48 Posts

    451

    0

    Posted May 5th, 2011 08:00

    It's showing the end of the dedup process:

    2011-05-05 14:49:58: DEDUPE: 6: PE: Task 1 finished scan on fsid 172, duration: 429327 seconds
    2011-05-05 14:50:08: DEDUPE: 6: 5: The deduplication scan on file system id 172 has just completed.
    So it took almost 5 days to scan this 2.5TB fs.
  • BillW

    1 Message

    451

    0

    Posted June 28th, 2011 12:00

    The F-RDE Policy Engine does not actually know how many real files (vs directories, links, etc) are on the filesystem when it starts the scan so it estimates the value based on the number of used inodes. Since this value may be somewhat inaccurate and we do not want to exceed 100% complete or hit 100% too early, it stops at 95% until the scan is complete.