Fix monitor Running status and CPU accuracy#128
Merged
Conversation
Two monitor accuracy fixes for worker heartbeats: Running status: get_tasks reserved a concurrency-semaphore permit *before* the blocking broker poll (a 5s XREADGROUP), and the heartbeat reported task_limit minus free permits as "in flight". An idle worker sitting in the poll therefore reported one task running even though it was only checking for work. Introduce a TaskSlots limiter that separates slot reservation (backpressure around the poll) from dispatch (a task actually handed to a worker); the heartbeat now reports only dispatched-and-not-yet-finished tasks. Workers already call limiter.release() once per task, so every worker is covered without per-worker changes. CPU accuracy: cpu_pct was derived from /proc/self/stat, which counts only the parent process. The CPU-bound workers offload their heavy work to ProcessPoolExecutor children, whose CPU never appears in the parent's counters, so those workers reported near-zero CPU while pegging cores. Read cgroup-wide CPU accounting (cpu.stat usage_usec on v2, cpuacct.usage on v1) when running in a container so pool children are included; fall back to /proc/self on a bare host (where /sys/fs/cgroup is the whole machine). This is accurate under both local Docker Compose and Kubernetes. Also report cpu_count (cgroup-limit aware) so the top-style "% of a core" reading is interpretable in the dashboard. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016L2RdZQn1gJ1iHd5FN6Dwf
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Two monitor accuracy fixes for worker heartbeats:
Running status: get_tasks reserved a concurrency-semaphore permit before the blocking broker poll (a 5s XREADGROUP), and the heartbeat reported task_limit minus free permits as "in flight". An idle worker sitting in the poll therefore reported one task running even though it was only checking for work. Introduce a TaskSlots limiter that separates slot reservation (backpressure around the poll) from dispatch (a task actually handed to a worker); the heartbeat now reports only dispatched-and-not-yet-finished tasks. Workers already call limiter.release() once per task, so every worker is covered without per-worker changes.
CPU accuracy: cpu_pct was derived from /proc/self/stat, which counts only the parent process. The CPU-bound workers offload their heavy work to ProcessPoolExecutor children, whose CPU never appears in the parent's counters, so those workers reported near-zero CPU while pegging cores. Read cgroup-wide CPU accounting (cpu.stat usage_usec on v2, cpuacct.usage on v1) when running in a container so pool children are included; fall back to /proc/self on a bare host (where /sys/fs/cgroup is the whole machine). This is accurate under both local Docker Compose and Kubernetes. Also report cpu_count (cgroup-limit aware) so the top-style "% of a core" reading is interpretable in the dashboard.
Claude-Session: https://claude.ai/code/session_016L2RdZQn1gJ1iHd5FN6Dwf