Page MenuHomePhabricator

Increase the limit of VCPUs for the recommendation-api project
Closed, InvalidPublic

Description

The recommendation-api project is at VCPU capacity and I'd like to request an increase of the number of allowed instances VCPUs. Ideally I could terminate some of the instances, but I need time to learn what each instance is doing. I don't have documentation regarding instances from the previous developer. Could you increase the number of CPUs from 24 to 36?

Event Timeline

bmansurov renamed this task from Increase the limit of instances for the recommendation-api project to Increase the limit of VCPUs for the recommendation-api project.Jan 12 2018, 4:28 PM
bmansurov updated the task description. (Show Details)
bmansurov added a subscriber: bd808.

@bd808 I'd appreciate your input to this task. This is a blocker for T183043. Thank you.

aborrero triaged this task as Medium priority.Jan 12 2018, 5:21 PM
aborrero subscribed.

We usually discuss quota modification request during our team meetings, the next one being 2018-01-16.

There are two xlarge instances in that project with very low current CPU usage. One seems to be using a lot of disk as a postgres server (experimental.recommendation-api.eqiad.wmflabs). The other (related-articles.recommendation-api.eqiad.wmflabs) does not have the extra disk quota mounted at all.

I think that experimental.recommendation-api.eqiad.wmflabs was related to T169766: Request custom instance for recommendation-api labs project. That request ended with a statement that the whole project was on hold, so maybe you can recover those resources?

I need time to learn what each instance is doing

Can we change this request to a temporary increase and set a deadline for determining how many resources can be given back to the global pool?

Would appreciate it if the author could convert the request to use the format at https://phabricator.wikimedia.org/project/view/2880/

Thanks for replies. I'll convert the request to use the right format. I may even be able to use one of the existing machines as Bryan suggests.

I was able to re-use one of the existing intances. I'll close this task as invalid. Thanks again, all.