We want to find a way to send ourselves an alert (probably by email) when a publishing run fails. At present, we mostly find out when an editor asks why a new version of a page is not available, or when Rhythmyx hangs because the temp folder is full.
Actually creating an alert is no problem; the issue is the trigger for this to happen. Since Rhythmyx usually hangs, we could run a process at the end of the run to send an “OK” confirmation, but if the run fails, the warning would never be sent. At present, our best idea is to monitor for the absence of an “OK” confirmation within a reasonable time of a run starting. However, runs are of very variable length, so this is not going to be very accurate.
Obviously we can use the publishing runtime panel to watch runs, but there is not always time to do that.
I think drossall was trying to say that the alert would never be sent if the server hung or the publishing of the edition hung…
You’ll either have to monitor the server.log with something like sec (simple event correlator) http://simple-evcorr.sourceforge.net/, or you’ll need to run some kind of poll on the database. The table of interest would be the PSX_PUBLICATION_STATUS. If an edition has started, but not yet completed, the START_DATE will be populated, but the END_DATE will be NULL. You could make a comparison of the current time with the start_date and determine if the edition appears to be hung. Of course, you’d have to determine what span of time constitutes a hung publishing run alert.
James is probably referring to the scheduling functionality that was added to Percussion CM in Version 6.6. A scheduled task for running editions is available as part of this functionality. You can define notifications to inform you of the results of these tasks, including errors that occur during publishing. In the Administration Manual (Version 6.7 link), see “Maintaining Schedules”, p. 21.
But can it notify when hung? My impression is that, once a process hangs, or runs very slowly, nothing more happens, including notifications. If it does manage to complete, you’ll get notifications then, but that could be hours late.
I want to be the first to know if a run is not going to complete.
vtdarrell seems to agree with me that Rhythmyx cannot do this. I’m not clear that rljohnson and jimbo are of the same mind.
If the server is hung, pretty much by definition it cannot inform you of its state. For that particular case, the scheduling functionality is not the answer.
I accept that a hung server cannot send reports to us. However, at present the issue is one of incremental runs taking far longer than they should (3-7 hours instead of 1-1.5).
Using vtdarrell’s principles, the CMS might allow us to set a warning time in minutes on each run, after which the server would notify us, presumably by email, that a run had not yet finished, so that we could take an active interest.
In the mean time, we’ll have to try an external program as vtdarrell suggests.
It’s not really reasonable to try to watch the publishing runtime tab during every publishing run, of course.
