[Gluster-users] Self-heals gone wild

[Jamie Lawrence]

Hello,

I recently stood up a 3x2 (soon to be 3x3) distribute-replicate volume on 5.9, running on Centos 7.7.

Volume Name: test_stage1_shared
Type: Distributed-Replicate
Volume ID: 99674d15-7dce-480e-b642-eaf7da72c1a1
Status: Started
Snapshot Count: 0
Number of Bricks: 2 x 3 = 6
Transport-type: tcp
Bricks:
Brick1: sc5-storage-1:/gluster-bricks/pool-1/test_stage1_shared
Brick2: sc5-storage-2:/gluster-bricks/pool-1/test_stage1_shared
Brick3: sc5-storage-3:/gluster-bricks/pool-1/test_stage1_shared
Brick4: sc5-storage-4:/gluster-bricks/pool-1/test_stage1_shared
Brick5: sc5-storage-5:/gluster-bricks/pool-1/test_stage1_shared
Brick6: sc5-storage-6:/gluster-bricks/pool-1/test_stage1_shared
Options Reconfigured:
features.quota-deem-statfs: on
nfs.ports-insecure: on
features.inode-quota: on
features.quota: on
server.allow-insecure: on
nfs.rpc-auth-allow: 10.181.51.190,10.181.43.10,10.181.70.190,10.181.70.191,10.181.70.192
transport.address-family: inet
nfs.disable: off
performance.client-io-threads: off


Moved data to it, and after a couple of days of letting it sit, came in this morning to a client /var filesystem full of:

[2019-10-08 17:34:31.249186] I [afr-self-heal-common.c:2868:afr_log_self_heal_completion_status] 0-test_stage1_shared-replicate-1:  metadata self heal  is successfully completed,   metadata self heal from source test_stage1_shared-client-3 to test_stage1_shared-client-4,  test_stage1_shared-client-5,  metadata - Pending matrix:  [ [ 0 0 0 ] [ 0 0 0 ] [ 0 0 0 ] ], on /sftp/wi/wcs/inprogress
[2019-10-08 17:34:31.260064] I [afr-self-heal-common.c:2868:afr_log_self_heal_completion_status] 0-test_stage1_shared-replicate-0:  metadata self heal  is successfully completed,   metadata self heal from source test_stage1_shared-client-0 to test_stage1_shared-client-1,  test_stage1_shared-client-2,  metadata - Pending matrix:  [ [ 0 0 0 ] [ 0 0 0 ] [ 0 0 0 ] ], on /sftp/wi/wcs/sdcom


On the server peers, I'm seeing things like:

[2019-10-06 12:30:04.636311] W [glusterd-locks.c:586:glusterd_mgmt_v3_lock] (-->/usr/lib64/glusterfs/5.9/xlator/mgmt/glusterd.so(+0xe5df0) [0x7f1a23e43df0] -->/usr/lib64/glusterfs/5.9/xlator/mgmt/glusterd.so(+0xe5d22) [0x7f1a23e43d22] -->/usr/lib64/glusterfs/5.9/xlator/mgmt/glusterd.so(+0xec2cd) [0x7f1a23e4a2cd] ) 0-management: Lock for test_stage1_shared held by e9fc174c-4156-475e-9d07-36b6c85a364f

But nothing else looks that unusual in those logs. I tried running a `gluster v heal` manually, which generated a ton of log noise but apparently little else.

Has anyone seen this before, or have a hint? I'm new to version 5.9, and have not run in to this problem before. 

Thanks,

-j
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4975 bytes
Desc: not available
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20191008/f9750958/attachment.p7s>