search

It4innovations / hyperqueue

Scheduler for sub-node tasks for HPC systems with batch scheduling
https://it4innovations.github.io/hyperqueue
MIT License
266 stars 20 forks source link

Distinguish hq-server folder to have multiple server for different machines share the same home #719

Closed unkcpz closed 5 days ago

unkcpz commented 5 days ago

In the supercomputer I use, it has two domains point to two type of machines. For user, the home is shared by two machine. That is to say, no matter which machine I use, I login and see the same home. But this cause the problem when I want to use hyperqueue on both machines. I can use --access-file to tell which hostname and ports to use, but only one hq-current. I start HQ server for one machine, the other one cannot connect to server. I guess client use info in hq-current to talk to server. Is there a way to make it independent to each other so I can run hyperqueue on both?

Kobzol commented 5 days ago

Hi, you can use the server directory parameter to run completely independent instances of HyperQueue.

To avoid having to pass --server-dir to all commands separately, you can also use the HQ_SERVER_DIR environment variable to set the server directory, and e.g. export it in the same terminal session:

$ export HQ_SERVER_DIR=/my/dir
$ hq server start &
$ hq worker start &
$ hq submit ...
unkcpz commented 5 days ago

Okay, thanks for fast reply! Will give it a try.