Port forwarding / tunneling

Opens a local server socket for clients to connect to, where traffic gets forwarded to a service on a HPC cluster login (or compute) node. This feature requires UNICORE 9.1.0 or later on the server side.

You can use this feature in two ways

in your own applications via the pyunicore.client.Job class.

you can also open a tunnel from the command line using the pyunicore.forwarder module

Here is an example for a command line tool invocation:

LOCAL_PORT=4322
JOB_URL=https://localhost:8080/DEMO-SITE/rest/core/jobs/some_job_id
REMOTE_PORT=8000
python3 -m pyunicore.forwarder --token <your_auth_token> \
   -L $LOCAL_PORT \
   $JOB_URL/forward-port?port=$REMOTE_PORT \

Your application can now connect to localhost:4322 but all traffic will be forwarded to port 8000 on the HPC login node where your application is running.

See

python3 -m pyunicore.forwarder --help

for all options.

If you want to tunnel to a compute node, you need to specify the compute node in your command line:

LOCAL_PORT=4322
JOB_URL=https://localhost:8080/DEMO-SITE/rest/core/jobs/some_job_id
REMOTE_PORT=8000
COMPUTE_NODE=cnode1234
python3 -m pyunicore.forwarder --token <your_auth_token> \
   -L $LOCAL_PORT \
   $JOB_URL/forward-port?port=$REMOTE_PORT?host=$COMPUTE_NODE \

If the remote UNICORE server is 10.2.1 or later, it is also supported to connect to a UNIX domain socket file on a login node. This is usually preferable for security reasons. The domain socket file must be in the UNICORE job working directory. To connect, simply specify the domain socket filename like so:

LOCAL_PORT=4322
SOCKET_FILE=sock
JOB_URL=https://localhost:8080/DEMO-SITE/rest/core/jobs/some_job_id
python3 -m pyunicore.forwarder --token <your_auth_token> \
   -L $LOCAL_PORT \
   $JOB_URL/forward-port?file=$SOCKET_FILE \

Again, this will only work if the job is running on a login node.