Closed
Bug 680572
Opened 14 years ago
Closed 13 years ago
[amo] Make code updates smoother
Categories
(mozilla.org Graveyard :: Server Operations: Projects, task)
Tracking
(Not tracked)
RESOLVED
FIXED
People
(Reporter: jbalogh, Assigned: oremj)
References
Details
Restarting the python processes to pick up new code gives us an awful load spike. It makes us scared to push updates when we're at high load and makes our graphs look bad.
My ideas:
* Take boxes out of the zeus pool (1-5(ish) at a time) and let them update in peace
* Try out gunicorn instead of mod_wsgi
Other than fear, this is the big blocker from us pushing more often.
| Assignee | ||
Updated•14 years ago
|
Assignee: server-ops → jeremy.orem+bugs
| Reporter | ||
Comment 1•14 years ago
|
||
Load: http://cl.ly/2S011N1x3w282a1V3m1J
Responses per second: http://cl.ly/1q1Q2c0P3g3n413p2P42
Comment 2•14 years ago
|
||
justdave uses the "few at a time" method for bugzilla updates, and he might have a script already to do that we could adapt to the deploy script.
but yeah, +1 to taking a couple of nodes out of zlb at a time and doing the updates.
| Assignee | ||
Comment 3•14 years ago
|
||
Yeah, I looked at that script a few months ago. Problem is, using that method with amo makes pushes take a really really long time, because the server count is so much higher and AMO has too much load to take large chunks out at a time.
| Assignee | ||
Comment 4•14 years ago
|
||
We won't attempt any fixes until after the PHX move.
| Assignee | ||
Updated•14 years ago
|
Component: Server Operations: Web Operations → Server Operations: Projects
| Assignee | ||
Comment 6•13 years ago
|
||
I don't think this is a problem anymore after the move to gunicorn.
Status: NEW → RESOLVED
Closed: 13 years ago
Resolution: --- → FIXED
Updated•10 years ago
|
Product: mozilla.org → mozilla.org Graveyard
You need to log in
before you can comment on or make changes to this bug.
Description
•