We recently updated the region of our app to
➜ ~ fly regions list -a staxcloud-staging
But the VMs still are in AMS, they are not changing:
staxcloud-staging at: 09:25:37
Name = staxcloud-staging
Owner = staxxer-848
Version = 136
Status = running
Hostname = staxcloud-staging.fly.dev
Platform = nomad
ID PROCESS VERSION REGION DESIRED STATUS HEALTH CHECKS RESTARTS CREATED
68b97c21 worker 136 ams run running 0 2022-09-03T12:25:14Z
a5678a8d web 136 ams run running 1 total, 1 passing 0 2022-09-03T12:24:55Z
c7dc6ec3 scheduler 136 ams run running 0 2022-09-03T12:24:15Z
I believe the region pool affects new vms. So you may need to deploy a new version and then those new vms should be started up in
fra. Assuming there is sufficient capacity etc. And the old
ams ones would then go away.
Thanks for your reply. This would make sense, but we’ve been deploying all day. Lots of new VMs have been created and they still get created in
I just tried this for another app:
fly regions set fra
and it immediately removed the old VM and creates a new VM in the new region
Ah … Strange.
If your other app is able to add its vms in
fra, doesn’t sound like a capacity issue either. Which would be my other guess for why new vms aren’t created in a particular region. Alas, one for Fly to debug.
This looks like a multi-process app. Do you have [processes] set in your config here? If so, you’d want to invoke
fly regions set fra --group $name
Right, this was it. So in what way does this behaviour make sense or is this just one of the rougher edges of the multi-process feature?
It could absolutely be better documented that
regions set takes a group argument !
I suppose it’s an extension of the flexibility in provisioning resources that you get with multi-process apps.
Wouldn’t it make more sense to see it as a filter? E.g.
fly regions set fra changes the region for all processes.
fly regions set fra --group web changes the region only for the
Yeah, that makes total sense, but I don’t have an ETA for this right now. We’ve recently been talking about ways to make using multi-process apps easier, and so I’ve brought this up in that conversation. Yesterday, we also added some validation in the API that will at least make it clearer when the
--groups flag is needed.