Can the server be deployed with high availability?

[Claude-Zq]

Currently, helm-chart only supports the deployment of a single server. Can multiple servers be deployed to achieve high availability?
https://www.jetbrains.com/help/teamcity/2024.07/multinode-setup.html?TeamCity%20Documentation

[sergelogvinov]

Hi, thank you for this information. I've been using TeamCity for more than six years and thought it was a single-daemon setup.

Yes, it makes sense to set it up as a high-availability solution... 👍

[Claude-Zq]

Will this helm-chart support high-availability deployment in the future?This is the only official helm I've found.
I found another helm-chart that supports HA, but it's not very active and, you know, unofficial.
https://artifacthub.io/packages/helm/teamcity-server/teamcity-server

[sergelogvinov]

Yes, I will add it to my list to implement. I have just one concern – it uses a shared folder between daemons, which could be a single point of failure. I don’t recall any downtime with TS, except during the upgrade process.

My chart is also unofficial, I do not work at jetbrains...

[Claude-Zq]

In my practice, the entire company needs to share a teamcity server instead of using one separately for each project, so a single teamcity server may become a performance bottleneck.

In addition, the network status between the server and the agent cannot be guaranteed.

So I need to test whether a high availability deployment can solve these problems

[TAMQuiroz]

@Claude-Zq When I was doing some tests with the latest chart version, I saw that it was possible to have the server with high availability by enabling Clustering Mode with 2 or more replicas. The load balancing worked when I manually changed the created headless service port to 80 and the ingress to point to the headless service because it pointed to the main pod. That way the requests come first through the headless service and then to any available pod, main or secondary.

@sergelogvinov will the headless service configuration be modified in the future? or maybe are you thinking of making it more flexible via chart?

[sergelogvinov]

Hi,

My first idea was to use first pod-0 to the humans requests (ingress), all others for the workers.
But if you have a better idea, I'm open to suggestions.

TeamCity ha currently implemented as an alpha version. There are many things to improve here )