Finally, the Linux cluster's up to 2.5.1. Out of the stone age, into the
bronze age! (My scientists are now researching Mathematics...)
In the process of upgrading, we noticed a few things.
First, the monitoring core RPM (at least on some mirrors ... ?) requires
librrd. I'm not sure if the metadaemon's being installed in all cases, but
I know it's not enabled by default when you run 'configure' ... I know
opening this can of worms complicates the RPM-building process, but it's a
needless dependency.
This is as of late last week, so I don't know if anyone's monkeyed with the
spec or rebuilt the RPMs since then.
Also ... rather puzzlingly, the front-end no longer seems to consider any
nodes as being down. I have traced this down to a problem in the released
ganglia.php:
if ($host_up)
$hosts[$cluster][$host] = $attrs;
else
- $hosts[$cluster][$host] = $attrs;
+ $hosts_down[$cluster][$host] = $attrs;
I'm sure Fed has caught this already in development but the
publicly-available tarball (and RPM?) out there may have it...
Hmm ... job view ... might have to update to CVS...
