These are chat archives for ManageIQ/manageiq/performance

2nd
Nov 2015
Keenan Brock
@kbrock
Nov 02 2015 15:12
@Fryguy selective builds ==> https://github.com/kbrock/johnny_five_sample
the idea is to validate that a build matrix even needs to be run
but it probably requires voice
Jason Frey
@Fryguy
Nov 02 2015 15:13
oh very interesting
Matthew Draper
@matthewd
Nov 02 2015 15:13
@kbrock I like the concept, but still couldn't work out what the config DSL was setting up
Keenan Brock
@kbrock
Nov 02 2015 15:13
right
need voice
it supports 2 dsls
component "ui" { |c| c.file "Gemfile" }
target (and dependencies)
OR
c.file "Gemfile", ["ui", "controllers", "models"]
so you specify dependency -> target
not sure which makes more sense
but it supports both :)
could move into yaml file pretty easily
@matthewd changed readme file
250 lines + 50 config == intimidating :(
Keenan Brock
@kbrock
Nov 02 2015 15:18
@matthewd all the logic: deduce
Jason Frey
@Fryguy
Nov 02 2015 15:25
Thanks @kbrock ...let's find a time, maybe today to have a bluejeans call
Keenan Brock
@kbrock
Nov 02 2015 15:27
that would be cool
Jason Frey
@Fryguy
Nov 02 2015 15:28
@kbrock Did you write JohnnyFive or is that an external library...it's hard to tell from the link you gave
Matthew Draper
@matthewd
Nov 02 2015 15:46
@kbrock sounds like you're going with something other than rubybench?
Alex Krzos
@akrzos
Nov 02 2015 16:17
So not sure if this is a first, but on my vmware-xlarge environment, the MiqEmsRefreshCoreWorker cycled several times during a Memory baseline test due to exceeding the default 400MiB threshold
Whats the relationship/purpose of the Core Worker vs the actual worker?
Keenan Brock
@kbrock
Nov 02 2015 16:18
@fryguy sorry - mine
@matthewd want to do 2 things. rubybench = storage of values after each run (summary)
but we need to collect various timings of each run. that needs a time series database
Keenan Brock
@kbrock
Nov 02 2015 19:11
@jrafanie @fryguy @matthewd ugh - I'm futzing around in this johnny five code - anyone up for a quick checking so I can at least validate whether this is going to go anywhere?
Matthew Draper
@matthewd
Nov 02 2015 19:56
I definitely like the idea in principle… the example configs just seem very complicated to me. Maybe a less surgical division would have benefit, without feeling so imposing, then we could get more precise over time?
Jason Frey
@Fryguy
Nov 02 2015 23:05
I updated https://gist.github.com/Fryguy/ec432fb7ed07767f8491 with which specs have been FIXED by PRs...I can run an updated number grab if needed
I also add another summary of query counts by spec file
I'm consdering adding this to the actual test suite...I don't think it introduces much overhead, and I think I can limit it to only when --profile is passed to rspec, and then spit out the top 10 or top 20
Joe Rafaniello
@jrafanie
Nov 02 2015 23:06
I"m going to tackle the miq_report_spec.rb ones... I already did half of them, just need to finish them
Jason Frey
@Fryguy
Nov 02 2015 23:06
nice :+1:
wow yeah, that's an expensive one overall... number 11 in top queries per file
Joe Rafaniello
@jrafanie
Nov 02 2015 23:07
this one is also still awaiting UI team review: ManageIQ/manageiq#4581
that will drop the miq_product_features_spec.rb