Page MenuHomePhabricator

Request increased quota for reading-web-staging labs project
Closed, DeclinedPublic

Description

Project Name: reading-web-staging
Type of quota increase requested: cpu
Reason: This project is used for a variety of reading web apps - tools to aid with Scrum (stack), tools to help us test changes before merging (reading-web-staging 2+3) and various team member supported and active experiments from 2 team members (pushipedia, reading-web-research, trending, vicky). I need at least 2 cpus to add a new project there. We like to keep them all in the same space so that all team members know where to look when there is a problem.

Event Timeline

Jdlrobson created this task.Sep 2 2016, 6:13 PM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptSep 2 2016, 6:13 PM

stack should really be moved to tool labs I think. I'd be glad to work with someone (@Jhernandez?) to get it running on Tool Labs. Is it taking up a whole VM at the moment or just colocated with something else?

@bd808 is pretty small.

Also @phuedx can you confirm that we use reading-web-staging-2 for testing Central Notice? (I notice that's a large vm which is probably overkill but I'm guessing too late/not worth the hassle to change)

phuedx added a comment.Sep 5 2016, 7:39 AM

We can safely reading-web-staging-2.

@phuedx We can safely remove reading-web-staging-2?

stack

@bd808 stack is taking a full vm of the smallest size. I'm happy to move it somewhere else if needed.

The minimum requirements for the app right now are (I believe) node 4+ and npm 3+.

I think I can add a dist task to create a tarball to avoid npm in the server (I don't think there are native dependencies). That'd remove the need to have npm 3+ on the server.

From what I see toollabs runs node 0.10, which LTS maintenance is about to be ended in october.

I'd prefer not have to go through the trouble of making the app compatible with such an old version of node, but I'm happy to migrate to it if there's an option to run in v4 (current active LTS) or even better v6 (to be active LTS the 1st of October).

vicky

Another option can be to collapse vicky bot into the same VM as stack. They're both low resource usage so they should live together fine.

That'd give us 1 extra cpu

reading-web-research

I think this one can be safely migrated to a 1 core, or maybe even collapse with the other ones. Need to think about it.


If there's great need for the resources I can make a task to work it out and get to it as soon as I can. I think I'd go for collapsing the 3 VMs into one of 1 or 2 cores, which should give us 2 or 3 cpus free.

Thoughts?

@phuedx We can safely remove reading-web-staging-2?

Yes. Sorry.

I've triggered termination of reading-web-staging-2 per convo on irc with @phuedx

If there's something we need to test like CN we'll work to set it up in reading-web-staging-3 and update our staging instructions on mw.org.

bd808 added a comment.Sep 6 2016, 4:38 PM

stack

@bd808 stack is taking a full vm of the smallest size. I'm happy to move it somewhere else if needed.

The minimum requirements for the app right now are (I believe) node 4+ and npm 3+.

I think I can add a dist task to create a tarball to avoid npm in the server (I don't think there are native dependencies). That'd remove the need to have npm 3+ on the server.

From what I see toollabs runs node 0.10, which LTS maintenance is about to be ended in october.

I'd prefer not have to go through the trouble of making the app compatible with such an old version of node, but I'm happy to migrate to it if there's an option to run in v4 (current active LTS) or even better v6 (to be active LTS the 1st of October).

Big surprise, the wikitech documentation is not quite up to date (or at least a bit confusing). The new hotness for running webservices on the Tool Labs grid is Kubernetes. I have tools.hatjitsu running on a Jessie image under Kubernetes. This provides nodejs v4.4.6 and npm 1.4.21. We are not likely to setup base images that provide versions other than the ones in Debian Jessie's upstream apt repos unless the WMF starts maintaining alternate packages for production use.

To run a tool under this environment, use webservice --backend kubernetes nodejs start. An interactive shell that can be used to install NPM modules, etc is reachable via webservice --backend kubernetes shell after the service is up. Trying this out by creating a stack tool account and playing with the setup is probably the best way to determine if it's easy enough to switch over to. There is a Labs project that we can use to redirect https://stack.wmflabs.org/ to https://tools.wmflabs.org/stack/ after everything is working.

It seems like there was some cleanup here that rendered the quota bump unnecessary. Can I get confirmation on where this is at? If still needed can this need be phrased in terms of what size of new instances it relates to Type of quota increase requested: cpu. Such as: we need enough cores to spin up 2 new mediums etc

Jhernandez closed this task as Declined.Sep 15 2016, 11:35 AM

I'm going to decline it for now given we've made some space @chasemp

If somebody has something specific for the request please reopen