Repository navigation
Replace kamaji-etcd with aenix-io/etcd-operator - #95
Conversation
Andrei Kvapil (kvaps)
changed the title
25bb31c to
c5f1c85
Compare
Andrei Kvapil (kvaps)
changed the title
9669420 to
3a28029
Compare
Signed-off-by: Andrei Kvapil <[email protected]>
3a28029 to
a6d0f7c
Compare
| | etcdOperator.readinessProbe.periodSeconds | int | `10` | | | ||
| | etcdOperator.resources.limits.cpu | string | `"500m"` | | | ||
| | etcdOperator.resources.limits.memory | string | `"128Mi"` | | | ||
| | etcdOperator.resources.requests.cpu | string | `"100m"` | | |
There was a problem hiding this comment.
Why is the request CPU 100m? This is quite a lot, I think you should start with 20-40m, but you need to know how much etcd eats at least.
| | etcdOperator.readinessProbe.initialDelaySeconds | int | `5` | | | ||
| | etcdOperator.readinessProbe.periodSeconds | int | `10` | | | ||
| | etcdOperator.resources.limits.cpu | string | `"500m"` | | | ||
| | etcdOperator.resources.limits.memory | string | `"128Mi"` | | |
There was a problem hiding this comment.
The memory limit is very small, an average k8s cluster with 100 pods will force etcd to use up to 300MB per instance. I think it should be increased to 384-512MB
| | kubeRbacProxy.readinessProbe | object | `{}` | | | ||
| | kubeRbacProxy.resources.limits.cpu | string | `"500m"` | | | ||
| | kubeRbacProxy.resources.limits.memory | string | `"128Mi"` | | | ||
| | kubeRbacProxy.resources.requests.cpu | string | `"100m"` | | |
There was a problem hiding this comment.
This is quite a lot, I think you should start with 20-40m.
| | kubeRbacProxy.livenessProbe | object | `{}` | | | ||
| | kubeRbacProxy.readinessProbe | object | `{}` | | | ||
| | kubeRbacProxy.resources.limits.cpu | string | `"500m"` | | | ||
| | kubeRbacProxy.resources.limits.memory | string | `"128Mi"` | | |
There was a problem hiding this comment.
It's hard to say here, but it's worth reviewing the limit, you don't really want to catch OOM
|
Marian Koreniuk (@themoriarti), there are defaults provided by the etcd-operator maintainers. We can consider other values, but I think it might be better to address this in the upstream project. |
|
We are waiting for a response from etcd-operator cozystack/etcd-operator#199 , if they make changes it will simplify the integration. |
| resources: | ||
| limits: | ||
| cpu: 500m | ||
| memory: 128Mi |
There was a problem hiding this comment.
need more memory up to 512MB
| cpu: 500m | ||
| memory: 128Mi | ||
| requests: | ||
| cpu: 100m |
There was a problem hiding this comment.
for start etcd not need more then 40m CPU reservation
|
Andrei Kvapil (@kvaps) Can you add the default value of requests and resource limits when creating an etсd cluster? |
Marian Koreniuk (themoriarti)
left a comment
Marian Koreniuk (themoriarti)
left a comment
There was a problem hiding this comment.
At the moment, we have gone through all the discussions and have come to the decision not to set any additional limits, but after the release we will monitor and return to the discussion of defaults in the next discussions. Fly fly our good etсd.
2 participants
This change replaces kamaji-etcd with our own etcd-operator, which provides a robust environment for running etcd clusters.
The previous kamaji-etcd has not been removed yet, so it will continue to operate for compatibility purposes.
fixes: #33