Migrate Farallon Staging hub to this repository. #379
Conversation
|
Getting closer!
Now at: |
for Makes sense it expires after 18h! |
|
#380 fixes the timeouts. Turns out if you don't have a directory named https://pilot-hubs.2i2c.org/en/latest/topic/storage-layer.html has more info about how we do storage and why this matters. |
Great! I was going to suggest looking at mount targets and security groups that sometimes generate timeouts at the time to mount EFS. |
|
@damianavila for the expiry of the generated kubeconfig, I see two options:
Actually, we can probably do both - do (2) first and follow up with (1) |
|
I've done (2), and opened #381 to track (1) |
It makes sense to me in the current context. |
Not complete yet
- Needs to be nested under singleuser - Explicitly set mem_limit too - otherwise, the default 1G limit of memory stays in, and k8s doesn't allow a limit to be smaller than a guarantee
EFS doesn't let you set ver=4.2
We mount this into all daskhub user accounts regardless - by default they don't do anything
Needs to be automated eventually
With KUBECONFIG=secrets/farallon.yaml kops export kubecfg --admin=730h farallon-2i2c.k8s.local
Referencing base-hub from helm charts templates is basically impossible - go templates can not use '-' in their name! This makes referencing them easier
yuvipanda
changed the title
|
Thanks for all the review, @damianavila! |
kopsproduces kubeconfig files we can check-in to the repo,and use to authenticate to
kopsclusters on AWS.We test this by moving the farallon staging hub to be deployed
from here, for #368
TODO:
kopsdoesn't work after about 12h