review Tecken instance size and production configuration
Categories
(Tecken :: General, task, P2)
Tracking
(Not tracked)
People
(Reporter: willkg, Unassigned)
References
(Blocks 1 open bug)
Details
Since the increase in sym file sizes, Tecken has been falling over in production (bug #1793984). It's probably time for us to re-examine the instance size and other configuration.
|
Reporter | |
Comment 1•2 years ago
|
||
I started a dashboard for analysis.
|
|
Reporter | |
Updated•2 years ago
|
|
|
Reporter | |
Comment 2•1 year ago
|
||
This should get done for the GCP migration because it informs pod sizing. I didn't get very far. Everything we know so far is in the Tecken GCP migration doc.
I'm unassigning myself so that it's back in the pool so someone else can pick it up.
|
|
Reporter | |
Comment 3•1 year ago
|
||
We're looking at switching the instance type from c5.large to c5d.large which switches the disk from EBS to local. The theory being maybe Tecken instances using enough disk iops during upload API bursts that Tecken instances are getting their disk iops throttled. That's covered in obs-68.
I'm going to WONTFIX this issue. Instead of going through the exercise of sizing the instance for AWS, we'll base the initial pod sizes on the AWS instance size and then iterate from there in GCP.
Description
•