These are chat archives for ManageIQ/manageiq/performance

2nd
Mar 2016
Keenan Brock
@kbrock
Mar 02 2016 01:25
select services.*, parse_tree.ancestry
from services
join (
WITH RECURSIVE parse_tree (id, ancestry) AS (
  SELECT parent.id, parent.id::text
  FROM services AS parent
  WHERE service_id is null
UNION ALL
  SELECT child.id, parent.ancestry || '.' || child.id::text
  FROM parse_tree parent
  JOIN services AS child ON child.service_id = parent.id
)
SELECT id, ancestry
FROM parse_tree
) AS parse_tree on services.id = parse_tree.id
order by parse_tree.ancestry;
Keenan Brock
@kbrock
Mar 02 2016 14:42
Feel like I'm following :girl: / :rabbit: / :cat2: into wonderland but in terms of fun, this is a 7.5/10
Alex Krzos
@akrzos
Mar 02 2016 15:24
@kbrock Was there anything different you had to do to get that db working on 5.4?
Keenan Brock
@kbrock
Mar 02 2016 15:24
I was speaking with @dmetzger57 about this
I'm running 5.4.z (dev env)
which is different
but besides creating a GUID, REGION, v2_key, it was pretty basic
the db is 25G
Alex Krzos
@akrzos
Mar 02 2016 15:25
ok, i have 50GB of disk space for the db
Keenan Brock
@kbrock
Mar 02 2016 15:25
the "My Services" tab is like a cartisian product with a fibanaci thrown in there - so there is room for improvement ;)
Alex Krzos
@akrzos
Mar 02 2016 15:26
i'll give it a new go, do I need to original GUID, REGION and v2_key, I thought I could just create a brand spanking new appliance, initialize the db, then rake evm:kill, dropdb vmdb_production, createdb vmdb_production, pgdump,then run the auth fix
Keenan Brock
@kbrock
Mar 02 2016 15:26
yea
I picked a GUID, set REGION to 999 and ran fix_auth
not sure what I did, because it failed one time. but don't think I did anything different the second go
so +1 to your approach
Alex Krzos
@akrzos
Mar 02 2016 15:28
ok so i'm not totally insane then
Keenan Brock
@kbrock
Mar 02 2016 15:28
no
Alex Krzos
@akrzos
Mar 02 2016 15:28
phew
Keenan Brock
@kbrock
Mar 02 2016 15:28
both your approach seems right
AND dennis is having all sorts of problems - probably on take #3
I'm not running in an appliance
@akrzos so you are not insane on either side :)
Alex Krzos
@akrzos
Mar 02 2016 17:04
oh man, I know why my appliance was jacked, stupid region is different this time
grrr
Alex Krzos
@akrzos
Mar 02 2016 17:11
yup this page is slooooow
Alex Krzos
@akrzos
Mar 02 2016 17:29
Completed 200 OK in 97558.8ms (Views: 627.2ms | ActiveRecord: 9959.5ms)
Keenan Brock
@kbrock
Mar 02 2016 17:29
:)
Alex Krzos
@akrzos
Mar 02 2016 17:29
@kbrock @dmetzger57 Do we have a patch to test?
Keenan Brock
@kbrock
Mar 02 2016 17:30
not yet
Alex Krzos
@akrzos
Mar 02 2016 17:44
gotcha, is there anything for me to test at this point with respect to that db? I can gather some baseline numbers and post those into the bz for the time being
Jared Deubel
@jdeubel
Mar 02 2016 19:34
@kbrock @dmetzger57 do you guys need any help in testing with that DB ?
You can see the slowness right?
Keenan Brock
@kbrock
Mar 02 2016 19:39
yes
I've made services page ~40% faster. looking at a few other areas to see if there is anything else I can do
still have a number of N+1 queries
Jared Deubel
@jdeubel
Mar 02 2016 20:10
ok
@kbrock Nice... what timing are you seeing now?
Keenan Brock
@kbrock
Mar 02 2016 20:17
was 75-78s. now 33-55s (in dev mode - so who knows)
hope the 33 is the real number
but... you know?
never know
still making over 10k queries
that is a lot of network / garbage
@jdeubel ^
Jared Deubel
@jdeubel
Mar 02 2016 22:15
Ok cool
Do you think it would be benefical to let them test that? At this point if they can shave off 20 seconds that is big for them especial if they know we are still working on it @kbrock
They are very desperate
Keenan Brock
@kbrock
Mar 02 2016 22:16
I'm still digging
I'll see if I can package this up
Jared Deubel
@jdeubel
Mar 02 2016 22:24
@kbrock Thanks... I make it known that this is not final fix .. and that we are still digging deeper but this should help them improve the timings