pt-stalk purge_samples slows down checks

[Gavin Towey]

currently in pt-stalk, the purge_samples function is called between each check iteration. It's meant to clean up old oprofile samples. However there are a couple issues with this:

1. If you're running pt-stalk as a daemon, you may end up collecting a lot of oprofile data. The default retention time of 30 days leaves a lot of data.
2. When the data piles up, the find command takes 3-5 seconds to run. This means if you're trying to set --interval=1 then you don't actually check that frequently, and stalk starts missing events.

I think the solution is to move the purge_samples call to happen only after the trigger has been hit, and a sample has been collected. There is really no need to check and delete files every iteration because no new files have been added! The continuous find command just wastes time/cpu/io.

I could submit a patch, but the change is trivial, just move the current call to purge_samples up about 10 lines or so. I could submit a patch if you like, but I don't think it's needed in this case.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PT-577