Page MenuHomePhabricator

Request increased quota for toolsbeta
Closed, ResolvedPublic


Project Name: toolsbeta
Type of quota increase requested: cpu & ram & instance count

Toolsbeta current quota usage is at:
Instances: 8/12
VCPUs: 19/20
RAM: 38GB/40GB
Floating IPs: 0/1
Security Groups: 2/10

Floating IPs should be good since the only one available should be enough. Only the bastion may need it (and hopefully the GSoC candidates can use ProxyCommand so even bastion won't need it)
Security Groups doesn't apply to communications within the project, right? It should be good in that case.

cpu & ram & instance are definitely not enough to setup the instances for T190893.

For the listed instances are are almost definitely needed:

  • toolsbeta-bastion-NN.toolsbeta.eqiad.wmflabs -- Most dev-ing and testing should happen here, so m1.medium
  • toolsbeta-k8s-etcd-NN.toolsbeta.eqiad.wmflabs -- m1.small in tools
  • toolsbeta-k8s-master-NN.toolsbeta.eqiad.wmflabs -- m1.medium in tools, toolsbeta paws master is also m1.medium
  • toolsbeta-flannel-etcd-NN.toolsbeta.eqiad.wmflabs -- m1.small in tools
  • toolsbeta-(grid-)?webgrid-generic-NNNN.toolsbeta.eqiad.wmflabs -- toolsbeta-grid-webgrid-lighttpd-1.toolsbeta.eqiad.wmflabs is m1.small
  • toolsbeta-worker-NNNN.toolsbeta.eqiad.wmflabs -- m1.large in tools, toolsbeta PAWS is also m1.large. Assuming at least m1.medium is needed
  • toolsbeta-proxy-NN.toolsbeta.eqiad.wmflabs -- active proxy tools-proxy-02, being m1.medium, has / used 6.6G, 15-min loadavg 0.91, mem 3462 / 3955 free, assuming m1.small needed

With the above flavors (3 m1.medium, 4 m1.small), this brings at least 7 additional instance, 10 VCPU & 20G RAM.

We currently have 3 m1.large instances (toolsbeta-paws-worker-100[1-3].toolsbeta.eqiad.wmflabs), Maybe we can release one or two of those and/or use a smaller flavor?

Event Timeline

zhuyifei1999 created this task.

We can use a smaller flavor for the toolsbeta-paws-worker-100[1-3], but I'm unsure if a m1.small will work as-is due to the storage allocated (removing the puppet mount class might solve this).

With some cleanup, we're at:
Instances: 3/12
VCPUs: 10/20
RAM: 20GB/40GB
Floating IPs: 0/1
Security Groups: 2/10

This brings us exactly within range of setting things up without increasing the quota. We can still reduce the flavors of (toolsbeta-paws-worker-100[1-2] to m1.medium, but I probably will not have the time for that soon, ideally we do this after T188912).

I still thing increasing the quota is a good idea, I can think of a number of ways we might want to create additional instances to test upgrades or tweaks to grid or k8s in the future. Doubling the quota sounds sane to me.

+1 and approved in team meeting

I'm a bit lost about what the actual ask is here. I've updated quotas like so:

cores: 30
ram: 61440
instances: 20

Is that enough? Too much?

zhuyifei1999 assigned this task to Andrew.

Should be enough. I'll create the instances and if that ends up way too small / large I'll reopen.