Hi @<1549564752575860736:profile|HappyTiger58> ,

is it possible to setup a services queue and an agent in service mode when using Kubernetese and how would one go about doing that?

There's actually no need to do that - the k8s glue agent is suitable both as a normal daemon agent and a services agent, it all depends on the specific k8s task pod template you define in the agent chart

From time to time when trying to acces to my projects, I'll get the following messages "Fetch Experiments failed" and "Fetch tags failed" and none of my projects and or tasks will be visible, I also might get the error message "Error 0 : You can't write against a read only replica." did anyone encounter the same issue and if so do you have any pointers on where to look to solve it?

This sounds like some database stability issue - @<1523701827080556544:profile|JuicyFox94> do you have any idea?

@<1523701827080556544:profile|JuicyFox94> Thank you for the pointers I'll look into our MongoDB setup!

Nice so if I understand correctly I should have a first k8s glue agent listening on the default queue that can execute tasks and I can also have a second k8s glue agent that listen to the services queue that will handle my pipelines? Also just to make sure that I setup everything correctly in the chart, should I modify the value "agentk8sglue.image.repository" or the value "agentk8sglue.defaultContainerImage" and what image should I use? Thank you!

You can’t write on readonly replica is about MongoDB. I guess you are using a multiple replica setup. In this case the mongodb dependency chart have a lot of parameters to tweak the system and maybe also an arbiter is good for you. But this is a huge topic regarding mongodb specific k8s setups.