Hi everyone,
I am trying to set up hsds in an azure VM that I host in the cloud. I used the VM that is offered on the Azure marketplace and ran the script. The docker containers seem to be running.
I installed h5pyd on the virtual machine to try out the CLI tools. Setting the server endpoint to the hsds_endpoint adress from the hsds/config/overrride.yml file. But I don’t seem to be getting a connection.
Also curl hsds_endpoint/about yields a connection time out (with the endpoint from the override.yml).
Some of my current questions:
Should the h5pyd CLI tools even work in the VM?
Is there a way to read the hsds_endpoint? For example from the Docker containers?
I saw that there is also an additional azure active directory setup: hsds/docs/azure_ad_setup.md at master · HDFGroup/hsds · GitHub
Does the python script in the VM perform this setup and would it mess with the way I am trying to access the server?
Hi - seems like Azure is popular today! (there was another post about HSDS on Azure…)
If you ssh to the VM, can you do: curl http://localhost:5101/about ?
If that works, get the public IP for the machine from the Azure Portal (the container itself doesn’t know what IP it may be reachable as) and try curl with the IP address. And if that works, check that you can access from a different machine. There are a couple of factors that can cause problems with the last step: the 5101 port not being open, or the machine being on a different VPC.
The hsds-_endpoint in the override.yml is not critical - it’s used to provide some references in the http responses, but won’t effect usage from h5pyd or the REST VOL. Once you’ve got the connectivity issues sorted out, you can go back and update the override.yml with the public IP address or dns name (if it has one).
The VM script can setup active directory authentication - but I’d recommend trying it out with the password file based authentication first and then tackling Active Directory.
I don’t recall at the moment if the h5pyd CLI tools are installed by default on the VM. If not, it’s easy enough to do a $pip install h5pyd and then run the CLI tools on the VM.
