Over in opensolaris-discuss, as an addition to a response to a question about sar data file retention, I went off about something else I noticed that's more SMF related (at least with respect to how the manifest for the service was written). So I'll put it here too, hoping someone will take a look at this. I'm pretty sure I can think of cases where the described behavior could cause problems...
Looking at the method script for the sar service (as of snv_97), I see that it creates the sys crontab file if it doesn't exist (for start) or deletes it unconditionally (for stop), directly, without using the crontab command. That might not work well if someone enabled the service when cron was already running (although if the file was created with the proper permissions, it might be ok if svc:/system/sar:default started before cron); once cron is running, it needs to be notified of changes to its files. Further, simply deleting the file is obnoxious insofar as it makes it a problem if someone wants to edit the sys crontab file, since their edits will be lost if the service is restarted (or on next boot). Ideally (IMO), the method script would check if cron (in the same zone) was running and choose accordingly between using crontab command (if it was) or directly modifying the files (if the service starts before cron). And if it's going to fool with that crontab at all, it should either do so more smartly (commenting out or un-commenting, but not simply deleting the crontab), or just leave it alone (people used to have to do that by hand, after all). Certainly if the service _does_ modify the crontab file, sar(1m) should say something about the scope of the changes to the sys crontab made by the svc:/system/sar:default service. Just because the current arrangement is more or less reasonable _if_ svc:/system/sar:default was enabled at boot _and_ nobody ever puts anything in the sys crontab (or even changes the collection intervals) doesn't make it reasonable outside of those limits. Personally, I think if I wanted it to maintain those lines, I'd have a template file separate from the method script that in SVR4 pkg terms was editable, so that it would survive package updates, and that had a comment on the end of the lines it supplied so they could easily be removed from the sys crontab without affecting other lines. -- This message posted from opensolaris.org
