Worker needs more power as the queue can get very long and we have to process it in a timely manner, to mitigate some slowdowns due to ressource exhaustion we can schedule the workers on a separate nodepool
We have an "SLA" Pool for those cases, this nodepool is basically an exclusive set of nodes to use which can be used by using Taints and Tolerations. This pool is not much used at the moment the nodes consist of the same config as the default ones (but they are less booked):
|Machine name||vCPUs1||Memory (GB)|
For this to work, we have to give a headsup to #technik before we use it as this will only work as long as we are aware of the deployment in order to have enough "room" for the deployment on the nodes.
- Could be partially solved by documentation and warnings in the README in https://git.panter.ch/catladder/gitlab-ci/
- At this point can probably not be enforced