malariagen / datalab

Repo for files and issues related to cloud deployment of JupyterHub.
MIT License
0 stars 1 forks source link

Change DNS on workers/pod nodes #57

Closed roamato closed 5 years ago

roamato commented 5 years ago

@slejdops (cc @alimanfoo)

I was trying to run the notebook in #56 but I can't load the data. The problem appears to be the end point as the pods are trying to talk to cog's "outside" IP addresses from "inside".

Workers should be using 172.18.255.1 and .2 and .3 for DNS servers and host cog.sanger.ac.uk should return:

cog.sanger.ac.uk has address 172.27.6.12
cog.sanger.ac.uk has address 172.27.6.15
cog.sanger.ac.uk has address 172.27.6.18

You can verify that it works if the following snippet runs quickly and with no errors:

import s3fs

storage_path = 'ag1000g-release/phase2/AR1/variation/main/zarr2/ag1000g.phase2.ar1'

s3 = s3fs.S3FileSystem(anon=True, client_kwargs=dict(region_name='us-east-1', endpoint_url="https://cog.sanger.ac.uk"))
store = s3fs.S3Map(root=storage_path, s3=s3, check=False)
roamato commented 5 years ago

I've been advised to have a look here for a Sanger specific configuration of k8s. Somewhere in there there are the specs for the DNS and whatever else it might be needed to make it work.

Also people are very responsive on the #helpme Slack channel, I suggest to start from there should you have any problem.

slejdops commented 5 years ago

Cool, thx

slejdops commented 5 years ago

I have changed the pod config to use 172.18.255.1 as the primary nameserver.

roamato commented 5 years ago

Thanks. I'm satisfied pods can see the correct end points, closing the issue.