Speed up resource model on large pod lists (skip client-side deserialization)#201
Open
karldebisschop wants to merge 1 commit into
Open
Conversation
- Parse the pod list from the raw API response (_preload_content=False + json.loads) instead of building the client's typed model objects, whose per-object deserialization dominates wall-clock time on large pod lists. nodeCollectData and main read plain dicts (camelCase keys); startedAt is reformatted from its RFC 3339 string. - Parse per-run config (tags, mappings, defaults, emoticon, config file) once in main and pass it to nodeCollectData instead of re-parsing it per node. - Add opt-in Exclude Namespaces (RD_CONFIG_EXCLUDE_NAMESPACES): exclude namespaces server-side via the field selector. Default empty (no change). - Add opt-in Use API Cache (RD_CONFIG_USE_CACHE): pass resource_version=0 so the apiserver serves the list from its watch cache. Default off. - Emit compact JSON (drop indent; keep sort_keys for stable output). Tests updated to dict-based pod fixtures.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
On clusters with many pods, the resource-model source spends most of its
wall-clock time in the Kubernetes client deserializing the list response into
typed model objects. This PR speeds that up and adds two opt-in options, with
no change to default behavior.
Changes
Skip client-side deserialization (the main win).
collect_pods_from_apinow requests the raw response (
_preload_content=False) and parses it withjson.loads(resp.data).get('items', [])instead of building the client'sper-object model tree, which dominates runtime on large pod lists.
nodeCollectDataandmainread plain dicts (camelCase keys), andstartedAtis reformatted from its RFC 3339 string.Parse per-run config once.
mainparses tags, mappings, defaults, theemoticon flag, and the config file a single time and passes them to
nodeCollectData, instead of re-parsing the same config strings for everycontainer.
Opt-in: Exclude Namespaces (
RD_CONFIG_EXCLUDE_NAMESPACES). Acomma-separated list of namespaces to exclude server-side via the field
selector (only when no specific Namespace is set). Defaults to empty, so
nothing is excluded unless configured.
Opt-in: Use API Cache (
RD_CONFIG_USE_CACHE). When enabled, passesresource_version=0so the apiserver serves the list from its in-memorywatch cache rather than a quorum read from etcd — faster on large clusters
and lighter on the control plane, at the cost of possibly-stale data.
Defaults to off (strong reads, unchanged).
Compact output. Drops
indentfrom the JSON output (keptsort_keysfor stable ordering), trimming output size.
Compatibility
Default behavior is unchanged: the same node attributes are produced, the
running/terminated filtering is identical, and both new options are off by
default. The unit tests were updated to build dict-based pod fixtures matching
the raw API JSON, with added coverage for the namespace-exclusion and
API-cache paths; the full suite passes.
Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com