[LU-2600] lustre metadata performance is very slow on zfs

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Affects Version/s: Lustre 2.4.0
• Labels: JL performance prz zfs

Environment

mdt is a zpool with 3 sata drives

  pool: pool2
 state: ONLINE
 scan: none requested
config:

        NAME STATE READ WRITE CKSUM
        pool2 ONLINE 0 0 0
          sdb ONLINE 0 0 0
          sdg ONLINE 0 0 0
          sdh ONLINE 0 0 0

• Severity: 3
• Assignee: Alex Zhuravlev
• Reporter: Minh Diep

Description

mds-survey show create and unlink are very slow

[root@mds01 mds01]# tests_str="create lookup destroy" thrlo=192 thrhi=192 file_count=3840000 mds-survey
Wed Jan 2 16:43:00 PST 2013 /usr/bin/mds-survey from mds01
mdt 1 file 3840000 dir 192 thr 192 create 2220.02 [ 0.00,19997.98] lookup 9429.79 [ 0.00,41998.40] destroy 1545.46 [ 0.00,15998.32]
done!

Issue Links

• is blocked by: LU-5041 FID Prefetching
• is blocked by: LU-7235 ZFS: dmu_object_alloc() serializes object creations
• is duplicated by: LU-2476 poor OST file creation rate performance with zfs backend
• is related to: LU-4108 Failure on test suite performance-sanity test_4
• is related to: LU-4968 Test failure sanity test_132: umount /mnt/ost2
• is related to: LU-4696 Test timeout on sanity test_51ba: nlink before: 70002, created before: 70000