[Gluster-users] NFS over gluster stops responding under write load

[Lonni J Friedman]

Greetings,
I'm running version 3.3.0 on Fedora16-x86_64.  I have two bricks setup
with a volume doing basic replication on an XFS formatted filesystem.
I've NFS mounted the volume on a 3rd system, and invoked bonnie++ to
write to the NFS mount point.  After a few minutes, I noticed that
bonnie++ didn't seem to be generating any more progress output, at
which point I started checking assorted logs to see if anything was
wrong.  At that point, I saw that the client system was no longer able
to write to the NFS mount point, and dmesg (and /var/log/messages) was
spewing these warnings like crazy (dozens/second):
nfs: server 10.31.99.166 not responding, still trying trying

Those warnings started at 14:40:58 on the client system, but oddly
stopped a few seconds later at 14:41:04.  Here's the full bonnie++
output (/mnt/gv0 is where the gluster file system is mounted as an NFS
client):
########
[root at cuda-ljf0 ~]# bonnie++ -d /mnt/gv0 -u root
Using uid:0, gid:0.
Writing a byte at a time...done
Writing intelligently...done
Rewriting...
########

Here's what's in the glusterfs logs at the moment:
##########
# tail etc-glusterfs-glusterd.vol.log
[2012-09-18 14:54:39.026557] I
[glusterd-handler.c:542:glusterd_req_ctx_create] 0-glusterd: Received
op from uuid: 1d3fb6c7-f5eb-42e9-b2bc-48bd3ed09e62
[2012-09-18 14:54:39.029463] I
[glusterd-handler.c:1417:glusterd_op_stage_send_resp] 0-glusterd:
Responded to stage, ret: 0
[2012-09-18 14:54:46.993426] I
[glusterd-volume-ops.c:492:glusterd_handle_cli_heal_volume]
0-management: Received heal vol req for volume gv0
[2012-09-18 14:54:46.993503] E [glusterd-utils.c:277:glusterd_lock]
0-glusterd: Unable to get lock for uuid:
e9ce949d-8521-4868-ad1b-860e0ffd8768, lock held by:
1d3fb6c7-f5eb-42e9-b2bc-48bd3ed09e62
[2012-09-18 14:54:46.993520] E
[glusterd-handler.c:453:glusterd_op_txn_begin] 0-management: Unable to
acquire local lock, ret: -1
[2012-09-18 14:55:47.175521] I
[glusterd-handler.c:860:glusterd_handle_cli_get_volume] 0-glusterd:
Received get vol req
[2012-09-18 14:55:47.181048] I
[glusterd-handler.c:860:glusterd_handle_cli_get_volume] 0-glusterd:
Received get vol req
[2012-09-18 14:55:49.306776] I
[glusterd-volume-ops.c:492:glusterd_handle_cli_heal_volume]
0-management: Received heal vol req for volume gv0
[2012-09-18 14:55:49.306834] E [glusterd-utils.c:277:glusterd_lock]
0-glusterd: Unable to get lock for uuid:
e9ce949d-8521-4868-ad1b-860e0ffd8768, lock held by:
1d3fb6c7-f5eb-42e9-b2bc-48bd3ed09e62
[2012-09-18 14:55:49.306844] E
[glusterd-handler.c:453:glusterd_op_txn_begin] 0-management: Unable to
acquire local lock, ret: -1
# tail -f cli.log
[2012-09-18 14:55:47.176824] I
[cli-rpc-ops.c:479:gf_cli3_1_get_volume_cbk] 0-cli: Received resp to
get vol: 0
[2012-09-18 14:55:47.180959] I
[cli-rpc-ops.c:732:gf_cli3_1_get_volume_cbk] 0-cli: Returning: 0
[2012-09-18 14:55:47.181128] I
[cli-rpc-ops.c:479:gf_cli3_1_get_volume_cbk] 0-cli: Received resp to
get vol: 0
[2012-09-18 14:55:47.181167] I
[cli-rpc-ops.c:732:gf_cli3_1_get_volume_cbk] 0-cli: Returning: 0
[2012-09-18 14:55:47.181214] I [input.c:46:cli_batch] 0-: Exiting with: 0
[2012-09-18 14:55:49.244795] W
[rpc-transport.c:174:rpc_transport_load] 0-rpc-transport: missing
'option transport-type'. defaulting to "socket"
[2012-09-18 14:55:49.307054] I
[cli-rpc-ops.c:5905:gf_cli3_1_heal_volume_cbk] 0-cli: Received resp to
heal volume
[2012-09-18 14:55:49.307274] W [dict.c:2339:dict_unserialize]
(-->/usr/lib64/libgfrpc.so.0(rpc_clnt_notify+0xa5) [0x328ca10365]
(-->/usr/lib64/libgfrpc.so.0(rpc_clnt_handle_reply+0xa5)
[0x328ca0f965] (-->gluster(gf_cli3_1_heal_volume_cbk+0x1d4)
[0x4225e4]))) 0-dict: buf is null!
[2012-09-18 14:55:49.307289] E
[cli-rpc-ops.c:5930:gf_cli3_1_heal_volume_cbk] 0-: Unable to allocate
memory
[2012-09-18 14:55:49.307314] I [input.c:46:cli_batch] 0-: Exiting with: -1
##########

I'd be happy to provide more if someone requests something specific.

Not sure what other information to provide at this point, but here's
the basics of the gluster setup:
##########
# gluster volume info all

Volume Name: gv0
Type: Replicate
Volume ID: 200046fc-1b5f-460c-b54b-96932e31ed3c
Status: Started
Number of Bricks: 1 x 2 = 2
Transport-type: tcp
Bricks:
Brick1: 10.31.99.165:/mnt/sdb1
Brick2: 10.31.99.166:/mnt/sdb1
# gluster volume heal gv0 info
operation failed
##########

I just noticed that glusterfs seems to be rapidly heading towards OOM
territory.  The glusterfs daemon is currently consuming 90% of MEM
according to top.

thanks