Hello list,
We're using Ganglia at the University of Michigan to monitor cluster
nodes, and we found an issue with 3.0.2. When sending status
messages from gmond to gmetad, messages over ~66600 bytes would be
truncated and the trailing </GANGLIA> tag (among a few others at the
end) would be missing, and the gmetad host would mark that client as
missing.
We found the problem to be in version 0.9.5 of the Apache Portable
Runtime (APR) that shipped with Ganglia 3.0.2. Upgrading to the
newest APR (0.9.7) fixed the problem.
We used the following procedure to correct the problem on Mac OS X
Server 10.4.4 Buid 8G32:
1) untar the ganglia sources
2) cd into the ganglia-3.0.2/srclib directory
3) remove the 'apr' directory
4) download the 0.9.7 sources of apr into this directory
(ganglia-3.0.2/srclib)
5) untar the apr sources
6) rename the resulting apr-0.9.7 directory to apr (or create a symlink)
7) move up one directory to ganglia-3.0.2
8) build/install as normal
Hopefully this will be of assistance to anyone seeing a similar problem.
Chris Black
LSA-IT
University of Michigan
