Hello All-
I have been experimenting with the "gluster volume top" commands in the
search for performance bottlenecks. For some bricks fewer than list-cnt
files are listed for read and write speed tests, and for some bricks no
filenames are listed at all even though throughput figures are reported
for all the bricks. See below for example, which shows read-perf
results from two bricks on the same server.
Brick: bdan13.nerc-essc.ac.uk:/nemo/glusterfs
Throughput 51.20 MBps time 0.0000 secs
No entries in list
Brick: bdan13.nerc-essc.ac.uk:/nemo8/glusterfs
Throughput 128.00 MBps time 0.0000 secs
MBps Filename Time
==== ======== ====
0 ...ice_assm/1989/ORCA025-FOAM_y1989m12_gridT.nc 2012-01-04
11:44:12.71083
0 ...ice_assm/1989/ORCA025-FOAM_y1989m05_gridT.nc 2012-01-04
11:44:09.250759
This happens whatever the value of list-cnt. Does this indicate a
problem with those bricks, or is there a bug in the gluster "top" feature?
-Dan.
