These are chat archives for ManageIQ/manageiq/performance

20th
Nov 2017
Jason Frey
@Fryguy
Nov 20 2017 15:19
:+1:
Keenan Brock
@kbrock
Nov 20 2017 22:17
Wow - the concurrent_safe_batch.rbfor metrics is saving a 768 line sql statement for a single vm c&u
1 in 5 lines is 941 characters long (other 4 of 5 are ~10 lines long)
@Ladas ^
Nick LaMuro
@NickLaMuro
Nov 20 2017 22:29
heh, I am actually looking at the file now myself
Keenan Brock
@kbrock
Nov 20 2017 22:30
so I removed the bottlenect in connect
and all of the sudden, perf_capture is looking pretty bad
Nick LaMuro
@NickLaMuro
Nov 20 2017 22:32
I am confused on so many levels... the main one being "I removed a bottleneck, now this is 'bad'"
I would expect "bad a$$" at least...
Keenan Brock
@kbrock
Nov 20 2017 22:32
lol
I cut the time by 20%
but I rebased
now it is 20x slower
Nick LaMuro
@NickLaMuro
Nov 20 2017 22:33
heh, I wish I could say that has never happened to me as well...
Keenan Brock
@kbrock
Nov 20 2017 22:33
(I did change the ems that I was pointing to)
Nick LaMuro
@NickLaMuro
Nov 20 2017 22:33
sounds like you didn't reproduce things properly
NickLaMuro @NickLaMuro pushes snobby glasses up on his nose a bit
Keenan Brock
@kbrock
Nov 20 2017 22:34
lol
Nick LaMuro
@NickLaMuro
Nov 20 2017 22:34
:neckbeard:
Keenan Brock
@kbrock
Nov 20 2017 22:34
it is ManageIQ/manageiq#16505
Nick LaMuro
@NickLaMuro
Nov 20 2017 22:36
so, the reason I was looking into concurrent_safe_batch.rb was because I actually came from saver/sql_helper.rb
Keenan Brock
@kbrock
Nov 20 2017 22:36
yea
Nick LaMuro
@NickLaMuro
Nov 20 2017 22:36
but looking that over, and I think Ladas has had the same idea before, that code could really be moved out of the graph refresh and be made more generic
Keenan Brock
@kbrock
Nov 20 2017 22:37
wait
Nick LaMuro
@NickLaMuro
Nov 20 2017 22:37
looking at Storage#perf_capture, the process_perfs section could really benefit from that
Keenan Brock
@kbrock
Nov 20 2017 22:38
yea