You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This also affects when asking for releases in a specific namespace. For example, helm tiller only shows the first 3 releases
# Helm list return 27 releases
helm list -a --namespace miguel | wc -l
28
# But helm list only the first ones
curl 'http://localhost:8080/api/tiller-deploy/v1/namespaces/miguel/releases'
{
"data": [
{
"releaseName": "api-d16636",
"version": "0.1.0",
"namespace": "miguel"
},
{
"releaseName": "awesome-hyena",
"version": "0.1.0",
"namespace": "miguel"
},
{
"releaseName": "crazy-ladybird",
"version": "1.0",
"namespace": "miguel",
"icon": "https://engineering.bitnami.com/images/projects/bitnami-logo.png"
}
]
}
# as you can see if we retrieve only the first 3 elements from helm list
# Only showing the first
helm list -a --namespace miguel | head -4
NAME REVISION UPDATED STATUS CHART NAMESPACE
api-d16636 1 Mon Jun 19 16:11:04 2017 DELETED artifacts-0.1.0 miguel
awesome-hyena 1 Mon Jul 17 13:26:40 2017 DELETED toolchain-web-0.1.0 miguel
crazy-ladybird 1 Thu Jul 19 10:36:55 2018 DELETED fabian-jasper-test-1.0 miguel
This has a set of consequences once the tiller instance reaches the 512 limit.
We only show the charts that are withing that 512 window, regardless of if we are asking for a namespace or not. We should double check the get release function since there must be a way to get a release information without relying on listing all. Actually the helm list --namespace foo seems to work as expected but not our listing.
Any new deployments will be shown as not found even though the deployment is successful.
I think that the scenario in which a cluster will contain more than 512 charts is not that likely but there are some low hanging fruit that we could fix.
If possible:
Re-implement listReleases(namespace) to not to rely on listing all elements in the namespace. helm list does it properly so there might be a way.
Re-implement get method to not to rely on the global list.
We might also want to increase the page size, helm CLI has a max option.
The text was updated successfully, but these errors were encountered:
I have noticed a couple of issues with the Helm proxy, all of them related to the fact of having a number of releases over 512.
For example, I have a cluster with 862 releases (including deleted)
But our proxy returns 512
This also affects when asking for releases in a specific namespace. For example, helm tiller only shows the first 3 releases
This has a set of consequences once the tiller instance reaches the 512 limit.
helm list --namespace foo
seems to work as expected but not our listing.I think that the scenario in which a cluster will contain more than 512 charts is not that likely but there are some low hanging fruit that we could fix.
If possible:
The text was updated successfully, but these errors were encountered: