search

puppetlabs / gatling-puppet-load-test

34 stars 44 forks source link

What to do with future improvements #408

Open RandellP opened 4 years ago

RandellP commented 4 years ago

Like we get postgres process stats, but they are mostly meaningless. We really need to look at the /proc//smap files of all the subprocesses (and main process) and sum up the pss to get an accurate measure of memory used. Prolly the same for cpu...

And since orch increases memory usage over time (days) until it stabilized, we should try to find a way to prime it so that it is stable at the start of scale testing, or restart it as well as part of scale testing to avoid it artificially reducing capacity. Similar with the database getting loaded up slowing things down.

And the NMT ticket: https://tickets.puppetlabs.com/browse/SLV-773

billclaytor commented 4 years ago

@billclaytor

Ticket: restart orch with puppetserver to avoid increased memory usage (required maint epic)

New epics: NMT (move ticket above into it) Postgres

billclaytor commented 4 years ago

In general, many smaller specific epics (even if they're below the line) is preferable to misc buckets of unrelated tickets.