HTCondor Project List Archives



[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-devel] Collectors not advertising themselves



The Collectors advertise themselves to the UW Collector as well as other collectors in the pool (and thus show up in condor_status -any), but take COLLECTOR_UPDATE_INTERVAL seconds to show up. This is because a collector will not advertise itself unless it knows about at least 1 startd in its hash table. The end result is that the collector in a pool will not appear in condor_status until, at the earliest, COLLECTOR_UPDATE_INTERVAL (which defaults to 15 minutes) since at startup the collector will not know about any other condor daemons yet. The offending code indicates this was done intentionally to prevent collectors with no startds from reporting to UW. The offending code:

    // compute machine information
    machinesTotal = 0;
    machinesUnclaimed = 0;
    machinesClaimed = 0;
    machinesOwner = 0;
        ustatsAccum.Reset( );
    if (!collector.walkHashTable (STARTD_AD, reportMiniStartdScanFunc)) {
dprintf (D_ALWAYS, "Error making collector ad (startd scan) \n");
    }

    // If we don't have any machines, then bail out. You oftentimes
    // see people run a collector on each macnine in their pool. Duh.
    if(machinesTotal == 0) {
                return 1;
        }

Is there (still) a problem with administrators setting up condor with a collector on each machine?

Rob