I’m not sure it’s an issue of configuration. If anything I’m thinking maybe it’s a performance issue.
I can regularly get a Jupyter instance with 1 core and 4 GB of memory, so it’s hard to say it’s an issue with the compute node’s resources. I would imagine your Jupyter jobs have as much resources, if not more.
I feel like there’s something obvious that we’re missing maybe? Is the OnDemand server under high load?
OK - strange I would not have pinned this to some configuration issue. I’d ask if this is consistent. I.e., you can always get a Jupyter instance running and it always runs smoothly.
All right, I will do some testing today and see if it is solved for real. There was one other change on the server at the time, where and admin set up ood.cori.ncsa.cloud to redirect to localhost.
ood.cori.ncsa.cloud {
import tls
route {
respond /robots.txt 200 {
body "User-agent: *
Disallow: /"
close
}
reverse_proxy https://127.0.0.1:18080 {
transport http {
tls_insecure_skip_verify
}
}
}
}
That was done before I redid certs. But I pretty much kept everything in the ood_portal file the same. servername is currently still localhost. If I go to that address, it ends up on localhost:18080. I can’t think of anything else I did on this, but I’m not the only one with admin privileges.
I was in sudo su mode when I was looking at httpd -V.