Hi @<1653932438834909184:profile|LackadaisicalSwallow14> , can you add some information please? Are you using a self deployed server or the community? Has something changed recently? Is the issue occurring for all users or only specific ones?
Hi, ofcourse:
we're using the clearml website,
two things have changed recently - we've had a change in team lead, and after we contacted some one on your side, they've moved ownership of our companies workspace to another persons user. second thing changed is we've registered to Pro services.
these are some error logs and the config process stdout
Does the issue occur for all users in the workspace?
not sure, currently only tried through my user
I suspect you might be using some wrong configuration file or some wrong/old credentials. Can you try on a brand new environment please? I would maybe even suggest testing on a different machine/user.
I'd suggest using CLEARML_CONFIG_FILE to point directly to the config file. You can also use other env variables to "circumvent" the config file in order to debug - None
Like mentioned I've tried doing that,
O
I have not tried though to use the CLEARML_CONFIG_FILE env varfiable
but I've tried passing secret key and token directly as env variable as per the doc
which resulted in the error as seen in configerror2osenv.txt
Also please try a different station from scratch. Again I suspect something is misconfigured in your environment
I've tried the steps mentioned on two different machines, I don't mind rebooting the second machine and retrying the steps including the CLEARML_CONFIG_FILE env variable
an anomaly that occured passed the point where the workspace was transferred is that comapnies workspace overriden my personal user's workspace
meaning instead of having a personal workspace and a company worksapce as was before the transfer, my user currently only has single workspace
also, at the configuration page on the website, the tokens are updated as being used, when I attempt at running the experiments
I've also suspected as much. I've asked the guys check out the credentials starting with TX4PW3O (What you provided). They managed to use the credentials successfully with errors.
Therefor it is a configuration issue.
This the internal test made by the guys. Everything worked from a fresh properly configured station.
I'm assuminmg 10:57 is your guys attempt
but 19:15 is mine from yesterday
And what was the result from 19:15 yesterday? The 401 error? Please note that's a different set of credentials
the 401 error was with the second set of credentials, when used as OS variables,
before that I've also tried as config file but got the same error as on the first machine (configerror1)
another thing that seems weird: I looked at the configuration page from another user in our workspace, on their page when they look at the companies workspace credentials, you can only see a single (different) credential token created, are these pages per user? or if they're per workspace I would expect to see the same tokens on both users view of the workspace configurations
They're per user per workspace. So each user will have their own unique credentials for every different workspace
Ok, I've found the problem:
I would run the experiment in a docker container, the credentials from the machine were mounted correctly to the container at first to /root/clearml.conf, later in the experiment the dockerfile changed such to create a user for experiment to run in (to prevent later root perms problems.) which then probably caused the problem due to the user not getting the creds from /root/clearml.conf
this doesn't explain as to why the env variables didn't work though
but the probem seems to have beensolved now and it was on our end