> Hi Richard, > > > Anyone looking into adding probes to cron yet? > > I am working on 6718765 and have worked out two > alternative solutions, of which > one involves changing the queue structures. > Unfortunately, my sponsor for these > at Sun is currently busy with some higher priority > work, so we won't get ahead > until January. > > At any rate, being at it, writing a dtrace probe > would be a nice exercise I'd be > interested in doing, if time allows. >
Right; while the difference in symbol visibility is interesting, what I really want is a nice way to monitor cron. Seeing the queues breaks down the jobs it is running by cron (queue 'c') vs at/batch (usually 'a' and 'b', although one could use anything other than 'c', I suppose). And additional probes for job start/stop/errors would be more usable for watching current activity than trying to parse the cron log (which may well have been disabled anyway 'cause it can be a pig). BTW, I think cron should default queues d-z to njob=0, and refuse (or fail) jobs where njob is 0 for the queue. That would be so people couldn't circumvent limits on queues a and b by taking advantage of the nonzero defaults for d-z by just using at -q d (or anything from d to z, actually). Explicit queuedefs entries are an answer to the defaults at least, but few enough people realize that queuedefs exists, let alone that queues other than a-c are possible. Maybe it's also unfortunate that cron apparently lacks a _per_user_ job limit, so that a single user's jobs could be prevented from denying service to other users. The lack of controls such as those just encourages control freaks to use cron.allow to only allow cron access to trusted accounts as absolutely needed. Some finer-grained control ought to be available to allow cron access to be more or less safely granted to more accounts. -- This message posted from opensolaris.org _______________________________________________ opensolaris-code mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
