Kubernetes cloud agent skip upgrade

Completed

Hi there,
we are using the Kubernetes plugin to start cloud agents in our local k8s cluster. Everything works great, but we are having some issues with the startup time of the agents. Every time a new cloud agent starts, the server issues an upgrade request and we need to wait approximately 1-2 minutes until the agent is connected and available to pick up jobs from the queue. This behavior kind of breaks our automation and leads to delays.
We use a customized image built based on the jetbrains/teamcity-agent docker image, which is the same version as our TeamCity server.
Is there a way to overcome this agent upgrade and speed up the initialization process?

Thanks

 

3 comments
Comment actions Permalink

Hi Florin,

 

when the build agent starts, it checks with the server its own version and whether there are plugins that need to be updated before proceeding. With brand new images, particularly if you are using extra plugins, the agents will need to upgrade, leading to this process. The solution for it is usually just to load up the image, let it perform the upgrade process, then take a new image once the process has finished. Would that be an option for you?

0
Comment actions Permalink

Hi Denis,

We have an automated process to create the Docker images with our buildagent requirements based on the TC agent so if we need to register the agent with the server, the whole process becomes cumbersome. Isn't there a way to dynamically load the plugins at runtime?

0
Comment actions Permalink

Hi Florin, 

We have an internal automated process of creating docker images. Basically, every time server is updated, we create a new docker image with an up-to-date agent. This also means that agent needs to be recreated. 

Also, there's no way to dynamically load/update plugins on the agent side. A restart is required.

0

Please sign in to leave a comment.