fix(edge): pin cold-ship/control-channel decoupling + e2e-validate cold path - #500
Merged
Merged
Conversation
…ld path The edge cold-fragment shipper is gated SOLELY on cold.enabled and is already independent of the control channel (which only carries coordinated-sampling PrecomputeConfigSet updates) — so a static-config edge with cold.enabled: true and control_channel.enabled: false DOES ship its cold tier. Investigation across processor.go / cold_fragment_shipper.go / cold_spool.go / flush.go / ingest.go found no control-channel gate on shipping; there was nothing to remove. This change pins that decoupling so it can't regress, and validates the cold path end to end against the runnable gorilla-merger: - Add TestColdShipsWithControlChannelDisabled: drives the processor through its real lifecycle (Start -> ConsumeMetrics -> flushAll -> async ship) with cold.enabled + control_channel.enabled=false, asserts ctrlChan==nil, and verifies gzipped ASAPFRG1 fragments reach the HTTP sink and decode back to the ingested series. A regression that re-couples shipping to the control channel fails it. - Add a load-bearing INVARIANT comment at p.shipWorker.start() in Start(). E2E validated live on this machine (see datasets_eval/cold-path-e2e-validation.md): edge-encoder ASAPFRG1 fragments -> gzip POST /ingest/gorilla (200) -> pending L1 block -> queryable via Thanos StoreAPI (30/30 samples round-trip) -> compaction to L2 -> live object-store PUT to a local MinIO bucket (uploaded=1; block chunks+index+meta.json verified in-bucket). Remaining docker/compose-gated: the cross-service stack (edge collector + data-plane + thanos-query/store-gateway read leg) and the full Fig 7 cold-arm latency/recall numbers. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Jun 14, 2026
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.
Context
The cold-archive ship from the edge feeds the cold-fallback evaluation (Fig 7 cold arm). The concern: under static config (
cold.enabled: true,control_channel.enabled: false) cold fragments never ship, so MinIO/archive stays empty. PR #373 fixed the QUERY-side routing; this is the EDGE-side follow-up.Investigation result (honest)
The edge cold-fragment shipper is already decoupled from the control channel. Cold ingest (
ingest.go), cold flush+ship (flush.go flushAll/flushShardWarmCold), and ship-worker startup (processor.go Start) are all gated solely oncold.enabled; the control channel (startControlPlane) is an independent path that only appliesprecompute.PrecomputeConfigSetupdates to sketch aggregators. A grep for any conjunction of {ctrl/ControlChannel/control} with {cold/ship/frag} in non-test Go returns nothing. There was no control-channel gate to remove. The static configdeploy/mvp-multinode/configs/asap/asap-otel-agent-asapedge.yaml(cold on, ship_endpoint set, control_channel disabled) ships cold under exactly this code.Change (minimal, defensive — no behavior change)
TestColdShipsWithControlChannelDisabledpins the invariant: drives the processor through its real lifecycle (Start→ConsumeMetrics→flushAll→ async ship) withcold.enabled: true+control_channel.enabled: false, assertsctrlChan == nil, and verifies gzipped ASAPFRG1 fragments reach the HTTP sink and decode back to the ingested series. A regression that re-couples shipping to the control channel fails it.p.shipWorker.start()inStart().E2E validated LIVE on this machine
Against the runnable
gorilla-merger(builtgo build ./cmd/gorilla-merger), using the sameasap-gorilla-goencoder the edge uses:POST /ingest/gorilla→200(30-sample XOR fragment)Seriesreturns the series, all 30/30 samples decoded backshipper uploaded blocks uploaded=1; blockchunks/index/meta.jsonverified in-bucket (numSamples: 30,thanos.labels: {cluster, merger, tier})Details:
datasets_eval/cold-path-e2e-validation.md.Still docker/compose-gated (NOT claimed)
The cross-service compose stack as one unit (edge collector + data-plane + thanos-query/store-gateway read leg) and the full Fig 7 cold-arm latency/recall numbers.
Tests
asapedgeprocessorgo test ./...— PASS (incl. the new test)gorilla-mergergo test ./internal/...— PASS🤖 Generated with Claude Code