buttercup icon indicating copy to clipboard operation
buttercup copied to clipboard

re-deployments of buttercup break Signoz login

Open ret2libc opened this issue 4 months ago • 4 comments

When you deploy Buttercup the first time, with a local Signoz deployment, and you open it in your browser you are asked to register an admin account to see the data. Then, if you undeploy the system and you re-deploy, when you try to open again SigNoz in your Browser, you get an error and you are automatically redirected to http://localhost:33301/something-went-wrong.

Image

This is probably because your browser still has the cookies of the old deployment, however that doesn't work anymore.

The workaround for now is to delete the cookies associated with http://localhost:33301 and reload the page at http://localhost:33301.

ret2libc avatar Aug 21 '25 14:08 ret2libc

@srikanthccv since you mentioned it, you might be interested in having a look

ret2libc avatar Aug 21 '25 14:08 ret2libc

I can take this one. Could you please assign it to me. Looks like a session cookie issue when redeploying. Should be fixable by clearing SigNoz session data during make undeploy or adjusting cookie config.

ArsalanAnwer0 avatar Sep 29 '25 16:09 ArsalanAnwer0

I've examined this issue and have a question before proceeding. I can add explicit PVC cleanup during make undeploy (including Redis session storage) and a user notification about clearing browser cookies. This addresses the Kubernetes side, but the browser cookie issue still requires manual clearing by the user. Would a PR with these changes be useful, or are you looking for a more complete automated solution?

I'm unable to fully test this without API keys for the complete environment setup.

ArsalanAnwer0 avatar Sep 30 '25 02:09 ArsalanAnwer0

I'm unable to fully test this without API keys for the complete environment setup.

What do you mean by this? To make the whole system work you only need one single key, like OpenAI API Key.

I've examined this issue and have a question before proceeding. I can add explicit PVC cleanup during make undeploy (including Redis session storage) and a user notification about clearing browser cookies. This addresses the Kubernetes side, but the browser cookie issue still requires manual clearing by the user. Would a PR with these changes be useful, or are you looking for a more complete automated solution?

If that's the only available solution, then it's better than nothing! Thanks for looking into this

ret2libc avatar Sep 30 '25 06:09 ret2libc