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
We're using Tekton Operator Pruner to delete old PipelineRuns from the cluster. Tekton Operator Pruner uses Tekton CLI to delete PipelineRuns. However, it's impossible to delete PipelineRuns both by their age and count in the namespace.
Allow using PipelineRun delete arguments "keep" and "keepSince" together.
Use case
Tekton runs PipelineRuns in multiple, dynamically created namespaces, some namespaces get more traffic, others - less (for understanding: production and development branch namespace).
Setting only keep-since argument is not an option because more active namespaces will hold too many irrelevant (old) PipelineRuns, this is a problem because having too many PipelineRuns in the cluster is in general bad - it affects Tekton performance - high CPU usage for controller, slow WEB UI, from our observations - the acceptable amount of total PipelineRuns in the cluster is around 2k.
Setting only keep argument is not an option, because some less often-used namespaces will keep irrelevant (old) PipelineRuns until the threshold is reached and again the problem here is that old PipelineRuns from multiple namespaces with less traffic stack up and the total number of PipelineRuns is too high in the Cluster.
Setting any of these settings to a low value is not an option, because then still relevant information may be deleted.
Having the ability to set both of these at the same time would help for the mentioned problems: it would be possible to set "keep" to like 500 PipelineRuns (this solves the problem of high-traffic namespaces) and "keep-since" with age of like 14 days to remove irrelevant (old) PipelineRuns (that would be beneficial for namespaces with not so much activity).
The text was updated successfully, but these errors were encountered:
Feature request
We're using Tekton Operator Pruner to delete old PipelineRuns from the cluster. Tekton Operator Pruner uses Tekton CLI to delete PipelineRuns. However, it's impossible to delete PipelineRuns both by their age and count in the namespace.
Allow using
PipelineRun delete
arguments "keep" and "keepSince" together.Use case
Tekton runs PipelineRuns in multiple, dynamically created namespaces, some namespaces get more traffic, others - less (for understanding: production and development branch namespace).
keep-since
argument is not an option because more active namespaces will hold too many irrelevant (old) PipelineRuns, this is a problem because having too many PipelineRuns in the cluster is in general bad - it affects Tekton performance - high CPU usage for controller, slow WEB UI, from our observations - the acceptable amount of total PipelineRuns in the cluster is around 2k.keep
argument is not an option, because some less often-used namespaces will keep irrelevant (old) PipelineRuns until the threshold is reached and again the problem here is that old PipelineRuns from multiple namespaces with less traffic stack up and the total number of PipelineRuns is too high in the Cluster.Setting any of these settings to a low value is not an option, because then still relevant information may be deleted.
Having the ability to set both of these at the same time would help for the mentioned problems: it would be possible to set "keep" to like 500 PipelineRuns (this solves the problem of high-traffic namespaces) and "keep-since" with age of like 14 days to remove irrelevant (old) PipelineRuns (that would be beneficial for namespaces with not so much activity).
The text was updated successfully, but these errors were encountered: