🐛 Compress components manifests if they don't fit in the configmap #192
Conversation
k8s-ci-robot
added
the
cncf-cla: yes
k8s-ci-robot
added
the
size/L
Fedosin
changed the title
Fedosin
changed the title
| // getComponentsData returns components data based on if it's compressed or not. | ||
| func getComponentsData(cm corev1.ConfigMap) (string, error) { | ||
| // Data is not compressed, return it immediately. | ||
| if cm.GetAnnotations()[compressedAnnotation] != "true" { |
There was a problem hiding this comment.
We should also add docs about this annotation, If users compress manifests manually and try to use them in an air-gapped scenario, this annotation has to be set on the configmap.
There was a problem hiding this comment.
Hey! I added the documentation, that describes this process for air-gapped environments.
k8s-ci-robot
added
the
needs-rebase
| return components, nil | ||
| } | ||
|
|
||
| // Otherwise we have to decompress the data first. |
There was a problem hiding this comment.
Q. to better understand this logic: what is the use case when it is already compressed and why we need to decompress in that case?
There was a problem hiding this comment.
In the normal workflow, when we create a provider for the first time, we download manifests from the provided URL and store them in a configmap to reuse later. Unfortunately, there is a size limit (1MiB) on the configmap in kubernetes, and for some providers we can't store there manifests directly.
Here is an example of capi provider that has infrastructure-components.yaml file exceeding 1MiB: https://github.com/oracle/cluster-api-provider-oci/releases/tag/v0.12.0
The error looks like this:
E0718 10:19:28.002488 1 controller.go:329] "msg"="Reconciler error" "error"="failed to create config map for provider \"oci\": ConfigMap \"infrastructure-oci-v0.12.0\" is invalid: []: Too long: must have at most 1048576 bytes" "InfrastructureProvider"={"name":"oci","namespace":"cluster-api"} "controller"="infrastructureprovider" "controllerGroup"="operator.cluster.x-k8s.io" "controllerKind"="InfrastructureProvider" "name"="oci" "namespace"="cluster-api" "reconcileID"="
3530ba70-285c-4a14-997b-5bdfc092c973
To workaround this we put compressed data in the configmap, which is 10 times smaller than the original. Later, when we generate clusterctl provider repository in the "load" phase, we decompress it and use as before.
k8s-ci-robot
removed
the
needs-rebase
By default we download capi provider manifests from a provided URL and store it in a confimap to reuse later. Currently there is a limit on the configmap size, which is 1mb. If we the manifests size is bigger, then the installation process fails. To avoid it we will compress manifests if they can't be stored in the configmap directly.
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: alexander-demicev The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
approved
| componentsConfigMapKey = "components" | ||
| ) | ||
|
|
||
| var _ = Describe("Create and delete a provider with manifests that don't fit the configmap", func() { |
There was a problem hiding this comment.
can we also have a test where we supply an already compressed configmap? Can be done in a follow-up PR
| 2. Create a configmap manifest from the archived data | ||
|
|
||
| ```sh | ||
| kubectl create configmap v1.9.3 --namespace=capz-system --from-file=components=components.gz --from-file=metadata=metadata.yaml --dry-run=client -o yaml > configmap.yaml |
There was a problem hiding this comment.
Not a blocking one and can be tackled in a follow-up if found needed: v1.9.3 is probably CAPZ provider version in this case, might be better to have different name (capz-cm / capz-v1.9.3 as an examples) for cm instead of just version.
k8s-ci-robot
added
the
lgtm
|
LGTM label has been added. Git tree hash: 695657d88198b3dda42ec6fb1b0377d091761662
|
What this PR does / why we need it:
By default we download capi provider manifests from the provided URL and store it in a confimap to reuse later. Currently there is a limit on the configmap size, which is 1mb. If we the manifests size is bigger then the installation process fails.
To avoid it we will compress manifests if they can't be stored in the configmap directly.
Which issue(s) this PR fixes (optional, in
fixes #<issue number>(, fixes #<issue_number>, ...)format, will close the issue(s) when PR gets merged):Fixes #20