From 8fa08e172b496a349d2d0a9fbd362d5e597c541a Mon Sep 17 00:00:00 2001 From: Matthias Bertschy Date: Mon, 7 Sep 2026 14:47:48 +0200 Subject: [PATCH] experiment(ci): bump storage pod CPU request/limit to test starvation hypothesis Follow-up to #949/#950's residual component-tests flakiness investigation (15-17/30 failing even after kubescape/storage#397's sharded-writer fix). #950 showed that raising shard count 8->16 and SQLite pool 10->24 did NOT help (21/32 failed, arguably worse), ruling out shard/pool headroom as the cause and pointing at the storage pod's CPU quota (500m limit, unchanged across all experiments so far) as a more likely constraint: under a starved CPU quota, more concurrent write-path goroutines add scheduling overhead without proportional throughput gain. This is a diagnostic experiment only, isolated from any shard-count change (shards are no longer configurable - fixed at 8 via DefaultSingleWriterShards in kubescape/storage). Bumps storage pod requests.cpu 100m->500m and limits.cpu 500m->2000m in tests/chart/values.yaml to see whether the residual failures - and specifically the storage-side "context deadline exceeded" / "sqlite: step: interrupted" errors - go away with real CPU headroom. Co-Authored-By: Claude Sonnet 5 Claude-Session: https://claude.ai/code/session_01LCMGT6Po2tSr1VEDVrbbYd --- tests/chart/values.yaml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tests/chart/values.yaml b/tests/chart/values.yaml index 7222346016..88a3298613 100644 --- a/tests/chart/values.yaml +++ b/tests/chart/values.yaml @@ -44,10 +44,10 @@ storage: app.kubernetes.io/part-of: "kubescape-storage" resources: requests: - cpu: 100m + cpu: 500m memory: 400Mi limits: - cpu: 500m + cpu: 2000m memory: 1500Mi nodeAgent: # The component tests run the agent with the Go profiler on. This used to ride