These are chat archives for ManageIQ/manageiq/performance

24th
Aug 2015
Dennis Metzger
@dmetzger57
Aug 24 2015 16:00
@akrzos what version VC do you run for the simulator? I have a virtualized 5.5 appliance from a previous life, but I get a ton of errors when I try to enable the simulator on it.
Alex Krzos
@akrzos
Aug 24 2015 17:11
@dmetzger57 I have VC 5.5 with embedded postgres
Dennis Metzger
@dmetzger57
Aug 24 2015 18:36
@akrzos so a couple of my config files were corrupt, reasonable for the xml parser to be unhappy. I've got the vcsim running with it's defaults (32 hosts, 1024 VMs, etc.), now I'll bump up the config a little.
Alex Krzos
@akrzos
Aug 24 2015 18:37
@dmetzger57 awesome, I just kicked off the job to run your changes against a small/medium/large provider environments, i'll kick off against the xlarge tonight
This message was deleted
Jason Frey
@Fryguy
Aug 24 2015 18:38
you can edit old comments here, btw :D
just hit the up arrow
haha nice
Alex Krzos
@akrzos
Aug 24 2015 18:39
:)
Alex Krzos
@akrzos
Aug 24 2015 19:34
@dmetzger57 looking good so far, medium environment initial refresh showing 140-143s before patch, after patch 114-120s, large environment initial refresh showing 556s-585s before patch, after patch 350s-366s
Dennis Metzger
@dmetzger57
Aug 24 2015 19:45
@akrzos I like the trend.....
Matthew Draper
@matthewd
Aug 24 2015 20:00
@dmetzger57 out of curiosity, are these changes along the lines of my ‘bulk’ thought?
(I may have missed it, but don’t think I’ve seen a perf related PR that would allow me to answer that question for myself)
Dennis Metzger
@dmetzger57
Aug 24 2015 20:01
They are, essentially took your logic which was in the refresh mix-in code and applied it to the vmware code path.
Matthew Draper
@matthewd
Aug 24 2015 20:01
Just wondering because those numbers sound disappointing to me — implying there was a benefit, but not “enough”
Alex Krzos
@akrzos
Aug 24 2015 20:02
I expect the benefit to be much larger for the X-large provider
large == 3,000 vms
of course faster would make me happier
Matthew Draper
@matthewd
Aug 24 2015 20:09
@akrzos if you have the logs handy, what sort of time is now spent between those two lines, on the large run you’ve done?
akrzos @akrzos pulls logs
4m vs 18s
Matthew Draper
@matthewd
Aug 24 2015 20:18
Cool, that does sound like we might be a bit closer to linear
medium = 1000 vms?
Alex Krzos
@akrzos
Aug 24 2015 20:21
yes
small = 100 vms
Matthew Draper
@matthewd
Aug 24 2015 20:25
So at a linear 120s per 1000 VMs, 10000 VMs should theoretically be around 20 min. Which is less than 90, but… still a bit away from ideal :smile:
Alex Krzos
@akrzos
Aug 24 2015 21:09
FYI that was without a VIMBroker
I will re-run with vm borker once I get the x-large numbers