[Gluster-devel] 3.4.0alpha3 performances

[Emmanuel Dreyfus]

Emmanuel Dreyfus <manu at netbsd.org> wrote:

> gnusrc.tgz      63197997 bytes  5.92 MiB/s
> sharesrc.tgz    7655370 bytes   8.08 MiB/s
> src.tgz         201202018 bytes 200.23 KiB/s
> 
> On the latter, it stalls a lot. It took almost a minute before reporting
> the first block written.  There is nothing in client log. Any idea of
> what I should look at?

In fact it looks more like a transition: glusterfs starts with nice
performances, and at some point it drops. Once the transition is done,
everything gets slow. Even a ls in the volume's root, which has few entries,
takes around 30 seconds, with no improvement if retried:

$ time ls /gfs33
manu tmp  x
   38.59s real     0.00s user     0.01s system
$ time ls /gfs33 
manu tmp  x
   39.04s real     0.00s user     0.00s system

Volume seems healthy:

Status of volume: gfs33
Gluster process                                         Port    Online  Pid
------------------------------------------------------------------------------
Brick silo:/export/wd3a                                 49152   Y       19216
Brick hangar:/export/wd3a                               49152   Y       17325
Brick hotstuff:/export/wd1a                             49152   Y       16750
Brick hangar:/export/wd1a                               49153   Y       13881
NFS Server on localhost                                 N/A     N       N/A
Self-heal Daemon on localhost                           N/A     N       N/A
NFS Server on 7c642e2e-4a29-476d-88c4-4bb675d2232b      N/A     N       N/A
Self-heal Daemon on 7c642e2e-4a29-476d-88c4-4bb675d2232
b                                                       N/A     N       N/A
NFS Server on 5aefe733-e9d7-4011-8e23-6af0de297dd0      N/A     N       N/A
Self-heal Daemon on 5aefe733-e9d7-4011-8e23-6af0de297dd
0                                                       N/A     N       N/A

-- 
Emmanuel Dreyfus
http://hcpnet.free.fr/pubz
manu at netbsd.org