[Bugs] [Bug 1200457] gluster performance issue as data is added to volume. tar extraction of files goes from 1-minute on empty volume to 20-minutes on volume with 40TB.

bugzilla at redhat.com bugzilla at redhat.com
Fri Aug 26 12:16:04 UTC 2016


https://bugzilla.redhat.com/show_bug.cgi?id=1200457



--- Comment #25 from Brian Foster <bfoster at redhat.com> ---
(In reply to Pranith Kumar K from comment #23)
> Ran strace on the bricks for the fast and slow case, these are the 10
> maximum syscall latencies for all of the syscalls. I am seeing high
> latencies for fsync and stats in the slow case:
> 
> Normal case:
> 
...
> fsync
> 0.018422
> 0.018479
> 0.018490
> 0.018572
> 0.018582
> 0.033627
> 0.036283
> 0.047154
> 0.078602
> 0.695221
...
> lstat
> 0.002063
> 0.002094
> 0.003154
> 0.003503
> 0.003998
> 0.004202
> 0.004760
> 0.005640
> 0.008268
> 0.015741
...
> 
> Slow case:
...
> fsync
> 1.500296
> 1.503076
> 1.513502
> 1.516628
> 1.522290
> 1.523135
> 1.535945
> 1.546670
> 1.548597
> 1.590475
...
> lstat
> 1.529056
> 1.529124
> 1.530052
> 1.531486
> 1.531578
> 1.531743
> 1.533177
> 1.722707
> 1.931263
> 2.287300
...

Ok, did we learn anything about whether the number of calls is roughly similar
between the fast and slow runs? If so, the next best thing is probably to try
and dig into where the latency originates via the function graph tracer.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the assignee for the bug.


More information about the Bugs mailing list