[Gluster-devel] lockless lookup cause disk to be kicked out

lidi at perabytes.com lidi at perabytes.com
Mon Jan 12 09:51:32 UTC 2015


Hi Xavier:

   I have a problem with disperse, outline below.

   
Version-Release number of selected component (if applicable):
mainline and any version with disperse

How reproducible:


Steps to Reproduce:
1. create a disperse volume with 3 disks, mount it to /cluster2/test
2. mkdir -p /cluster2/a/b/c
3. on one console execute 'for ((;;));do dd if=/dev/zero of=/cluster2/test/a/b/c/aa bs=1M count=200;done'
4. on another execute 'ls /cluster2/test/a/b/c/aa' many times

Actual results:
one disk was kicked out



Additional info:
[2015-01-12 09:08:02.050709] W [ec-combine.c:801:ec_combine_check] 0-test-disperse-0: Mismatching xdata in answers of 'LOOKUP'
[2015-01-12 09:08:02.052371] W [ec-common.c:162:ec_check_status] 0-test-disperse-0: Operation failed on some subvolumes (up=7, mask=7, remaining=0, good=5, bad=2)
[2015-01-12 09:08:02.052479] I [ec-heal.c:546:ec_heal_init] 0-ec: Healing '/a/b/c/aa', gfid 658bbea7-c18d-48e5-9a54-21549bdb31eb
[2015-01-12 09:08:02.054985] W [ec-combine.c:76:ec_iatt_combine] 0-test-disperse-0: Failed to combine iatt (inode: 11120550027160400363-11120550027160400363, links: 1-1, uid: 0-0, gid: 0-0, rdev: 0-0, size: 104857600-0, mode: 100644-100644)
[2015-01-12 09:08:02.055086] N [ec-generic.c:819:ec_combine_lookup] 0-test-disperse-0: Mismatching iatt in answers of 'GF_FOP_LOOKUP'
[2015-01-12 09:08:32.741658] W [ec-combine.c:801:ec_combine_check] 0-test-disperse-0: Mismatching xdata in answers of 'LOOKUP'
[2015-01-12 09:08:32.742744] W [ec-combine.c:801:ec_combine_check] 0-test-disperse-0: Mismatching xdata in answers of 'LOOKUP'
[2015-01-12 09:08:32.742803] W [ec-common.c:162:ec_check_status] 0-test-disperse-0: Operation failed on some subvolumes (up=7, mask=7, remaining=0, good=6, bad=1)



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-devel/attachments/20150112/a866945d/attachment.html>


More information about the Gluster-devel mailing list