5 months ago
Is there any way to get a notification when a service is automatically redeployed by railway? Our services are getting redeployed, and then stop working because of version problems, etc… (for example a rabbitmq was automatically redeployed, and all services that use the rabbitmq stopped working) - it is extremely frustrating
I am not getting any notification about a redeployment so I usually catch the error after a few hours when I get the first complaints
0 Replies
Few more infos for the rabbitmq: I used the template and deployed it a few months ago - never had any issues. No with that auto-redeploy it stopped working for other services because there was a version mismatch? Is it possible that the redeployment also triggered an upgrade?
No, currently there is no advanced warning of any kind that your services might get redeployed.
As for your version mismatch - Railway will do exactly what you tell it to do with your predeploy and build commands. If your predeploy or build commands include updating your libraries then it would do so
yeah, I get that - but when I tell my services that they should upgrade on deployment, I would also appreciate the control to decide when to redeploy the service? Or at least get some insights that a service was redeployed automatically
I'm completely with you on this. This is frustrating.
Forced deployment - sure, it can be accommodated. But when it redeploys in the middle of the night with no warning, such that you couldn't have your engineers on-call to handle possible breakages - that's simply bad.
People do complain about this and Railway has a proven track record of implementing user-friendly updates to their operations. There is a possibility this will get better.
5 months ago
feel free to open a #🤗｜feedback thread
5 months ago
Hello,
As for why we redeploy sometimes -
As for why that has broken your RabbitMQ installation, likely two things, you have continued to use the :latest image version after deploying the template, and you are using an old version of the template, a version that did not implement a work around needed to keep RabbitMQ's data in the same static location within the volume, so every deploy, although RabbitMQ's data is not technically lost, RabbitMQ will pick a new location within the volume to save it's data, the latest version of the template fixes this oddity with RabbitMQ so that data is always saved in the same static location within the volume and thus you do not lose your RabbitMQ data (queues, etc).
5 months ago
!s
Status changed to Solved brody • 5 months ago