502, 500: BadGateway Error and Internal server error when Launching cvat

[bedanar]

Actions before raising this issue

• I searched the existing issues and did not find anything similar.
• I read/searched the docs

Steps to Reproduce

No response

Expected Behavior

I have installed cvat following the installation guide, my OS is Linux. After running the comman docker compose up -d the following happens in the Google Chrome:

Possible Solution

No response

Context

I want to launch cvat. Previously I already had problems with launching (it was 405 error), but it was successfully solved. Now for some unknown reson I have this problem.

Environment

No response

[blueclowd]

Also having the same 500 error, it seems like the internal server is not that stable.
Based on my observation, it usually happens when the server is idle for a while or right at the login stage.
It can be solved by rerun the container (may need to tried a few times) on my side
Also wondering how to prevent this problem.

[bedanar]

okay, thanks, I will try

[bsekachev]

Version?

If this is current develop branch, please pay attention that ENV_CVAT_REDIS_HOST was removed in this pull request: #7245

But it is still used in the container according to logs you provided. Did you try to rebuild docker images with docker compose up -d --build?

[bedanar]

I reinstalled the platform yesterday according to https://opencv.github.io/cvat/docs/administration/basics/installation/, I think it is the most recent version. I just did 'docker compose up -d --build', the error is still there.

[bsekachev]

@bedanar

Try:

docker compose down
git checkout v2.9.2
docker compose up -d

I suspect in your case it uses built docker images from the latest release, but compose config from develop branch,

[bsekachev]

Alternatively, you may use:
CVAT_VERSION=dev docker compose up -d to install develop version.

And one more suggestion is to built everything from sources:
docker compose -f docker-compose.yml -f docker-compose.dev.yml up -d --build

[bedanar]

Unfortunately, the error is still present after all these commands

[InRiPa]

Had the same issue as well. Wanted to try the compose variation again to maybe figure out, the reason that i can't get the helm chart alternative to run.

The proposed try

docker compose down
git checkout v2.9.2
docker compose up -d

worked for me.

Got it fixed through learning that docker system prune -a doesn't remove volumes anymore
Running (well aware, that I used it on a test system) docker volume prune --all resolved the issue for me.
Full steps:

docker compose down
git checkout develop
docker system prune -a
docker volume prune --all
docker compose up -d

Maybe it helps.

[bsekachev]

Removing docker volumes is not necessary, and it is dangerous. Creating risk that you remove all existing data.

[mpolinowski]

@InRiPa Thank you for sharing.

I was able to get past the cannot connect to the server issue using security_opt: seccomp:unconfined. But then I was running into status 504 - method not allowed errors when trying to log in.

Nuking everything ended up working.

This did not help - 504s remained:

docker compose down
git checkout v2.9.2
docker compose up -d

The following - as expected - got rid of all the data but left me with a so far functioning installation:

docker compose down
git checkout develop
docker system prune -a
docker volume prune --all
docker compose up -d

Now let's figure out how to restore projects from backups :)

++ Edit : Restore worked 👍🏻