We have been debugging a problem in 3.1.7 when using gmetric with
SPOOF_HOST on large clusters.  The symptom is that the individual host
data is correct for all hosts and the summary_info data is correct for
clusters with <32 hosts, but the summary_info rrd file is missing on
clusters with >32 hosts.
 
Latests TOT code fixes this problem.  It may be related to this fix in
process_xml.c:
 
 
> /* $Id: process_xml.c 2277 2010-02-05 22:03:45Z bnicholes $ */
6a7
> #include <ganglia.h>
810a812,818
> /* do not add every SPOOF_HOST element to the summary table.
> if the same metric is SPOOF'd on more than ~MAX_EXTRA_ELEMENTS hosts
> then its summary table is destroyed.
> */
> if ( strlen(new_name) == 10 && !strcasecmp(new_name, SPOOF_HOST) )
> return 0;
 
 
Anyway, we are now using the TOT version in production.  So far without
problems.  The improved calendar support is nice also.  This change to
support clusters with >32 hosts is helpful to us, hopefully it will be
retained in the next release.
 
If we have problems with the TOT snapshot, would it be possible for us
to patch this one change into a known good release?  Suggestions on best
path forward?
 
Thanks,
Ron
 
------------------------------------------------------------------------------
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general

Reply via email to