These are chat archives for ManageIQ/manageiq/performance

26th
Oct 2015
Dennis Metzger
@dmetzger57
Oct 26 2015 13:04
OK, fun morning trying to catch up with Fridays activities :smile: I went to pull the latest nightly build to test all the changes that have been going in for the memory issue, however it seems we've not had a good (complete) upstream nightly to pull down since the 16th, am I mistaken?
Alex Krzos
@akrzos
Oct 26 2015 13:20
@dmetzger57 Yeah I ran into that a few days ago, 16th was last good build I could find
Joe Rafaniello
@jrafanie
Oct 26 2015 13:27
Yeah, try git pull on both repos, @akrzos do you have those steps ?
Matthew Draper
@matthewd
Oct 26 2015 13:30
Step 3 is now bin/update, if that helps
Alex Krzos
@akrzos
Oct 26 2015 13:54
@jrafanie crud, I don't think I saved exactly what I did aside from bash history
akrzos @akrzos checks
Joe Rafaniello
@jrafanie
Oct 26 2015 14:02
i think the general idea was appliance && git pull && vmdb && git pull && bundle exec rake db:migrate
Matthew Draper
@matthewd
Oct 26 2015 14:03
appliance && git pull && vmdb && git pull && bin/update
Alex Krzos
@akrzos
Oct 26 2015 14:04
I have a few git stash / git stash apply as well
appliance was updated as well
Joe Rafaniello
@jrafanie
Oct 26 2015 14:05
keep in mind we have a few memory improvements in the works
Oleg Barenboim
@chessbyte
Oct 26 2015 14:05
@jrafanie any that will help my memory?
Joe Rafaniello
@jrafanie
Oct 26 2015 14:06
it’s ironic, I don’t remember what they were ;-)
just kidding, here’s one: ManageIQ/manageiq#5094
Matthew Draper
@matthewd
Oct 26 2015 19:13
I dropped the ball on trying to get a Rails 4.2 rc in time, so we should probably decide whether we want rc1 on 4-2-stable, or the current release
I have in fact started things in motion now, so we can still anticipate a new release sometime during our next sprint, before our final
Jason Frey
@Fryguy
Oct 26 2015 19:15
I wonder if we should lock to a SHA on 4-2 stable
I'm concerned that if we do any follow up builds it will automagically pull in any new changed and that may have unexpected issues
Matthew Draper
@matthewd
Oct 26 2015 19:16
Isn't that equally true of any gem's patch level releases we don't lock down?
Jason Frey
@Fryguy
Oct 26 2015 19:16
(also don't say Gemfile.lock ;) )
yeah, technically, but gems generally don't release commit by commit which is what tracking a branch does
unless 4-2-stable is infrequently committed to
Matthew Draper
@matthewd
Oct 26 2015 19:17
I was actually about to say that though ;)… maybe for the purpose of build follow-up/respin, it would indeed be worth hanging on to
Jason Frey
@Fryguy
Oct 26 2015 19:17
yeah, JoeV started down the path of storing the Gemfile.lock between builds, but it never got realized
Matthew Draper
@matthewd
Oct 26 2015 19:18
.. but it is a backbranch, so not quite the same level of adventure as following someone's master
Joe Rafaniello
@jrafanie
Oct 26 2015 19:18
not having a sha would make it hard to track down what version of rails is on the appliance... you'd have to deploy it and see what git sha upstream was at
not sure how'd you'd know downstream, hope for a downstream commit message with which git sha on 4-2-stable was used and/or look at the gems/bundler directory?
Jason Frey
@Fryguy
Oct 26 2015 19:44
well, you would know the SHA because bundler puts the SHA in it's file path
but that's kind of nasty to figure out during diagnostics
Joe Rafaniello
@jrafanie
Oct 26 2015 21:22
I found a "use less memory" thing in the refresh code...
Oleg Barenboim
@chessbyte
Oct 26 2015 21:22
nice! please make a PR
Joe Rafaniello
@jrafanie
Oct 26 2015 21:24
basically, we were updating our relationships table post refresh with the objects added/updated in the refresh... the objects we were sending to the relationship mixin was the full objects (Vm)... it turns out we only need the id of each, so we can do .select(:id) to query less things and create smaller objects
working on it @chessbyte ;-)
according to total_allocated_objects in GC.stat, it creates ~3 million less objects on the refresh of the large vsphere enviroment, I don't believe it's that much but who knows
Matthew Draper
@matthewd
Oct 26 2015 21:35
This message was deleted