ExasperatedCrocodile76 running_remotely() is a utility function you can import from the clearml package and does not require you to call Task.init() or initialize any clearml component. Use it as follows:
` from clearml.config import running_remotely

if running_remotely():
...
else:
... `

I would assume so, but you'll have to shut down the entire system to do that

Hi @<1600661428556009472:profile|HighCoyote66> , redis is not a concern, but the MongoDB is the main store of metadata and administrative data for the ClearML system. Did you delete the containers or the data folder? Is it possible this action corrupted the data folder?

Regarding the original message, base_url refers to the API endpoint (in a self-served server, this would usually be with port 8008). The fact you had to use /api simply means you were using the WebApp endpoint which has its own apo reverse proxy, which should not really be used by external clients (it's possible, but there's no guarantee for backwards compatibility there as it's designed to be used by the WebApp and might change in the future)

The UI uses it to access the list of users for presenting in the tasks table 🙂

I assume the next version is a matter of a couple of weeks

@<1561885921379356672:profile|GorgeousPuppy74> user credentials cannot be set from the configuration files - they are dynamically created (and can also be revoked by the user).

Hi ShaggySquirrel23 , if you run the experiment locally when you have a local git diff, ClearML will log this diff in the task (as well as the commit ID of the latest commit on which the diff has to be applied).
I'm not sure what I see in the log you attached - is this the output of clearml-agent? what is the exact flow you are using?

No no, the upload is done on the client side. This is for the server

You need to set that in the environment section of the apiserver service in the docker-compose.yaml file. And yes, you'll need to run docker-compose up again

ScantChimpanzee51 , one thought - If you take the s3 link used by the UI and try to download it using the ClearML SDK, does that workl?

Hi @<1535069219354316800:profile|PerplexedRaccoon19> , what you're missing is all the ElasticSearch data

That's not supported right now can it's a nice feature - we'll appreciate a GitHub issue on the matter 🙂

Well, it.might simply be the elasticsearch driver reusing connections. Regarding the apiserver, the CPU load is not indicative - how many requests per second, approximately?

If there are models used by other tasks, they will be kept there under the existing structure

ReassuredTiger98 we'll check, thanks (and sorry 😞 )

Hi @<1523713917677735936:profile|EnviousStarfish54> ,
In your experiment's Results/Scalars sub-panel you should see a graph titled :monitor:machine which contains a series per monitored value (e.g. CPU usage etc.)

When you restart the server, can you access it?

Can you attach the entire response (or preview)?

Got you 🙂 , thanks, we'll take a look

If so, you'll need to add a volume mount that will be applied to any docker container the agent spins up. You can do that by adding this in your clearml.conf file:
agent.extra_docker_arguments: ["-v", "/path/to/creds.json:/root/creds.json"]

@<1529271085315395584:profile|AmusedCat74> ?

Can you show me the browser's network tab when you try to login?

It looks like not all the containers are up... Try sudo docker ps and see if the apiserver container is restarting...

This also suits what you described - perhaps the server has been running for a long time and the system disk was filled by ES/files and perhaps logs?

Hi ExasperatedCrocodile76 , you can check the clearml function running_remotely()

Hi @<1715175986749771776:profile|FuzzySeaanemone21> , I assume this is because there's some issue with the SSL termination set up in the server

Hi @<1546665634195050496:profile|SolidGoose91> , this is something that should be added in one of the upcoming releases

So the issue is only SSH to the server, or the server not responding to the SDK and/or WebApp?

Since they won't be used, I guess you can just specify one package