Developers should lint the charts before committing and running.
On your local machine run (at root of the repository)
docker run -it --rm --name ct --volume $(pwd):/data quay.io/helmpack/chart-testing sh -c "cd /data; ct lint --config ct.yaml"
To run changes to the charts without a PR, you can login to Openshift, edit values-pr.yaml if needed and run (add a --dry-run flag for testing compilation of charts):
helm upgrade -f ./charts/traction/values.yaml -f ./charts/traction/values-pr.yaml pr-00-traction ./charts/traction --install --wait
We have 3 Openshift namespaces (dev, test, prod), each with a limit on resources. We have a 4th namespace (tools) with smaller limits (cpu: 4 cores -> 4000m).
limits.cpu = 8 cores -> 8000m limits.mem = 32Gi -> 32609Mi
requests.cpu = 4 cores -> 4000m requests.cpu = 16Gi -> 16304Mi
Note: tools namespace has reduced limits on cpu:
limits.cpu = 4 cores -> 4000m
requests.cpu = 2 cores -> 2000m
Currently, each instance has 3 containers (traction api & traction acapy & traction db). We are using the dev namespace for a DEV instance plus our Pull Request Instances (4 developers, so hopefully only 4 PRs open at a time). This puts cosiderable limitations on the resources we can request and our limits. We are leaving extra space for standing up tenant-ui on demand.
For test and prod namespaces, we can assume that we will need 2 instances (test + uat), (prod + hotfix).
For now, just keeping it simple by giving all containers similar requests and limits, this must be adjusted as we determine actual load and run requirements for each container.
Take the dev namespace limits divide by 6, then use a fudge factor of around 75% of max. The DEV instance (pre-test) will get double the resources than PRs that is why 6 instead of 5 (1 dev + 4 prs).
limits.cpu = 200m limits.mem = 820Mi
requests.cpu = 120m requests.cpu = 400Mi
See above, basically use double what a PR gets
limits.cpu = 400m limits.mem = 1600Mi
requests.cpu = 200m requests.cpu = 820Mi
TEST and PROD, expect 2 full instances then use a fudge factor of around 75% of max; give a little breathing room.
limits.cpu = 600m limits.mem = 2600Mi
requests.cpu = 300m requests.cpu = 1300Mi