Meteor Galaxy Random 502 Errors

We recently (< 24 hours) began experiencing random 502 errors on specific views on our app. This has never happened before so we are trying to process why this would randomly start occurring.

Ex URL => https://pocketlistautos.com/a/Qyd1RUtL

When visiting this URL, the URL loads ~50% of the time without issue. The other 50% the loading times out with a 502 error.

We have APM activated on our app but no material errors to report.

We have reached out to support but the response times are delayed. We are considering moving to another hosting provider but wanted to connect with the meteor community first.

Our database infrastructure provider is MongoDB Atlas with an M30 instance and upgraded IOPS.

Now our builds are randomly failing. We are on a 3-Container professional plan. No material changes to the code.

We have to investigate alternatives. Has anyone used Heroku? Others? Any recommendations for hosting?

I can highly recommend Convox - it’s provisioning resources for you on AWS ECS (or another cloud provider - but I’ve only used them with AWS) and the whole setup ends up being highly reliable and performant.

1 Like

Will check that out @brianlukoff.

Here is the error log for the new build failures:

v0180 2023-08-18 11:35:42-06:00Error trying to start deploy - out of Memory in the build worker, you can try to increase the total available RAM in your app settings here: https://cloud.meteor.com/customer/jeremypocketlist/push-to-deploy/pocketlist.autos.

How do we increase the total available RAM in our app if we are already on the Professional tier?

Guessing that this might be @denyhs area that he might know something about.

I need to mark this as resolved.

We discovered the issue after deploying the app to Heroku to cross-test: It was a promise (using the future package) to an external API that was hanging for > 2 mins and triggered the Meteor method 2 minute timeout. Once we setup better error handling the issue completely resolved itself.

In addition, we learned that when hosting on Galaxy a user can also increase the RAM available for builds outside of the hosting container settings. Very helpful! Once we were directed to that setting we were able to resolve the build issues as well.

Someday we would like to find some better logging. There were no logs on this issue and it was time consuming finding a resolution. We have also not tried APM services outside of Meteor APM.

Thanks for anyone who helped.

2 Likes