sorry for the delay ResponsiveRabbit81
unlike AWS where the UI would rely on key and secret to sign a S3 resource, in google storage, the UI expects a GS cookie, which is available after signing in to GS. are you signed in to GS in both Firefox and Chrome and can view the resource in GS console on both browsers?

@<1531445337942659072:profile|OddCentipede48> that's deliberate, URL contains experiments page preferences to give the user an easy method of sharing his current view. you can either update the link to the page with the updated choice or remove all table info from the URL (everything after the experiment ID) and they will be restored from the BE.

SillySealion58
something look odd with the bucket settings.
url has port 443 which usually means ssl secure, but the secure flag is false

Hi @<1523702496097210368:profile|ScantChimpanzee51> , currently we rely on a GCP token to be available on this session. in your case a private windows is a new separate session.
we consider signing GCP resources similar to AWS, in a future version

can you make sure in users table you can see the pending invite?
and on the invited account you see the invite under the user menu (top right corner)?

thanks, this issue might be related to github login, we're investigating

great

@<1533257411639382016:profile|RobustRat47> I was told that Github accounts are private by default, and won't pass the user's email to the newly created user in clear.ml
I'm not sure if logging in with a different provider like google will update the clear.ml user email, but it's worth a try to logout and login again with another SSO provider

@<1572757572547514368:profile|GiddySnake50> do debug sample images stored in that bucket show up correctly on the debug samples tab?
I'm trying to understand if S3 assets signing works correctly.

Hi @<1572757572547514368:profile|GiddySnake50> , where is this asset saved? fileserver? S3? GS?

the bug is during creation of S3 bucket credential creation. you can manually fix the URL in settings -> Configuration
under WEB APP CLOUD ACCESS find the right Host (Endpoint) field and add the missing colon.
we will release a fix to 1.9 soon.

thanks 😊

I was just informed that the images for 1.9.2 were already released on Thursday, please test if the new version solve the issue for you and let us know.

it might be a bug we already fixed, let me check

FYI we use the port to distinguish between AWS and local S3

DepravedSheep68 did you try curl http://localhost:8080 from the instance?

also, https won't work, it's simple http

if you get a response while on the instance from localhost, but not from the outside, it usually means the machine IP is not exposed to the outside world.

can you please also report in github so we have an item to follow once the fix gets released?
https://github.com/allegroai/clearml-web/issues

credentials moved to workspace tab, but I get the feeling your API server is down

does the webserver instance fails to run due to this error?

Thanks JitteryCoyote63 , we'll try to find the time for the effort to improve this behavior.
in the mean time, maybe open a ticket here https://github.com/allegroai/clearml-web/issues so maybe someone from the community might help.

Hi, maybe this is related None ?

OK thanks, looking into this.

Hi @<1705029129084080128:profile|SmoothPigeon36>
we found out that unintentionally, debug sample will display up to 3 images with the same iteration/metric/variant. which was not the initial intent. while image viewer support 1 image per this combination.
I have 2 questions.

• can you confirm that this is indeed the issue in your case.
• and if so, was reporting more than 1 image for the same combination intentional?

thank you @<1705029129084080128:profile|SmoothPigeon36> glad to hear you enjoy ClearML.
I believe we identified the source of the problem, and working on a bug fix.

there is no limitation reporting the same images on different iterations. what we think might be the issue that multiple images are reported with the same iteration + metric + variant, using report_image
can this be the case?

adding a running index to the variant name should take care of this issue

Hi EnviousStarfish54 ,
what version of server and SDK are you using?
can you report this experiment to demo server?

Hi FunnyTurkey96
can you please tell what was in the error "More Info"?
and if there are any relevant errors in the browser console, open it with F12, and select the console tab.