When looking at my dtlog today. I noticed that snmpdiskspace.monitor, ( the official release, and ours ) dumps the header line instead of the failed hosts into (last_summary) if -list or -listall was specified.
In the process of fixing that, ... What follows applies to our release only. ... I found that some of my hosts where not reporting partitions although all the snmp data was there. In the process of investigating that, I found this and wanted to bounce it off the list before releasing it. In the hostmib processing code right after # ignore this instance # and try to move on # to next we wouldn't # need this if # use-dummy-values # really worked If I add $instancenum = $instancenum - 1; I can recover from the error, and get the data. What I found is that, Its not usually the current instance of the dataset that is corrupted. Usually a previous dataset was incomplete, which skews the instance #'s. If we decrement the instance #, we can recover the current dataset. Using that one line fix, however could cause a show stopper. If the previous incomplete dataset's instance # was numerically right before the current instance#. Decrementing would cause code to loop infinitely. I tested this by setting the Instance# to one that was causing us problems, instead of decrementing. Here, decrementing the instance# works fine. The question is what is the risk of it breaking stuff elsewhere. -- Sincerely, Nathan Gibbs Systems Administrator Christ Media http://www.cmpublishers.com
signature.asc
Description: OpenPGP digital signature
_______________________________________________ mon mailing list mon@linux.kernel.org http://linux.kernel.org/mailman/listinfo/mon