SWIP-15: implement BanyanDB self-observability (cluster / container / group model)#13903
Merged
Conversation
… group model) Rebuild otel-rules/banyandb around the cluster reality: Service = cluster, ServiceInstance = container (pod_name + container_name, with role/tier attributes), Endpoint = group. Add banyandb-endpoint.yaml; redesign service/instance rules to mirror the upstream FODC-proxy Grafana boards. Requires BanyanDB 0.11+. Rewrite the e2e to a no-FODC file-discovery cluster (1 liaison + 1 hot data node); the collector scrapes each node's :2121 directly and injects the identity labels. Operator docs rewritten to the cluster/container/group model. Validated: DSLClassGeneratorTest compiles all rules via the production path; the e2e passes 16/16 against BanyanDB 0.11. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
… hardcoded image) The BanyanDB self-observability rules need the 0.11 FODC/queue/lifecycle metric families, which post-date the previous pin (#1139). Bump SW_BANYANDB_COMMIT to the 0.11-dev build (8a1936ce9) and have the banyandb case inherit it via base-compose instead of pinning the image SHA per-case. Verified: standalone boots on the new build (the 88 standalone cases' path) and the banyandb so11y e2e passes 16/16. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
There was a problem hiding this comment.
Pull request overview
This PR rebuilds SkyWalking’s BanyanDB self-observability integration to match SWIP-15’s cluster / container / group model (BanyanDB 0.11+), updating MAL rule files, documentation, and the BanyanDB e2e case to validate service/instance/endpoint scopes against a minimal 2-node cluster.
Changes:
- Redesign BanyanDB MAL rules into service (cluster), instance (container), and endpoint (group) rule files, adding a new
banyandb-endpoint.yaml. - Rewrite the BanyanDB e2e case to run a BanyanDB 0.11+ cluster (liaison + hot data) with direct per-node Prometheus scraping and injected identity labels.
- Rewrite BanyanDB dashboard documentation and update
CHANGESto reflect the new entity model and metrics.
Reviewed changes
Copilot reviewed 12 out of 12 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
| test/e2e-v2/cases/banyandb/otel-collector-config.yaml | Scrape both BanyanDB nodes directly and inject identity labels for the SWIP-15 rule model. |
| test/e2e-v2/cases/banyandb/nodes.yaml | Add static file-based node discovery for the 0.11+ cluster e2e. |
| test/e2e-v2/cases/banyandb/expected/metrics-has-label-value.yml | New expectation template for labeled time series validation in e2e. |
| test/e2e-v2/cases/banyandb/e2e.yaml | Adjust failure cleanup to collect logs from the new multi-service layout. |
| test/e2e-v2/cases/banyandb/docker-compose.yml | Stand up a 2-node BanyanDB cluster (file discovery) and configure OAP to use it as storage. |
| test/e2e-v2/cases/banyandb/banyandb-cases.yaml | Update e2e queries to the new meter names and entity identities (cluster/instance/group). |
| oap-server/server-starter/src/main/resources/otel-rules/banyandb/banyandb-service.yaml | Redefine Service-scope metrics as cluster KPIs keyed by cluster. |
| oap-server/server-starter/src/main/resources/otel-rules/banyandb/banyandb-instance.yaml | Redefine Instance-scope metrics as container KPIs keyed by pod_name@container_name with role/tier attrs. |
| oap-server/server-starter/src/main/resources/otel-rules/banyandb/banyandb-endpoint.yaml | New Endpoint-scope (group) KPI rules keyed by group under the cluster service. |
| docs/en/changes/changes.md | Add changelog entry describing the SWIP-15 BanyanDB self-observability redesign. |
| docs/en/banyandb/dashboards-banyandb.md | Rewrite documentation to explain the cluster/container/group entity model and metric catalog. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
…se-compose BanyanDB 0.11 removed etcd-based coordination; the banyandb-data/liaison base service defaults still used --etcd-endpoints, which is an unknown flag on the 0.11 image (now the SW_BANYANDB_COMMIT pin). Switch the defaults to --node-discovery-mode=file (the convention every cluster case already overrides to). Verified --etcd-endpoints is the only 0.11-invalid flag across the e2e tree; all TLS/auth/cache-wait/node-discovery/data-node-selector flags remain valid. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…e windows) Address review feedback: query_latency, merge_file_latency/partitions, gc_pause_avg, and disk-usage-percent divided counters/rates with the `/` operator, which yields NaN/Infinity when the denominator rate is 0 (no queries / merges / GC events in a window). Switch every division to SampleFamily.safeDiv (returns 0.0 for an empty/zero denominator), matching the shipped envoy-ai-gateway latency rules. Boot-check (DSLClassGeneratorTest) recompiles all rules clean. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Member
Author
|
Thanks for the review. Addressed the division feedback: every ratio metric ( |
Second half of the query_latency review comment: dividing the raw cumulative liaison_grpc_total_latency / _started counters yields an all-time average, not a windowed latency. Rate() both operands (matching the upstream Grafana "Query Latency" panel) so the metric reflects recent latency; safeDiv already guards the idle window. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Member
Author
|
Re-checked the |
…ner/group model MALExpressionExecutionTest loads the rule expressions from the (rewritten) otel-rules files but pairs them with the committed .data.yaml mock input + expected output. The banyandb fixtures still carried the old single-node host_name model, so every rule ran against input with no `cluster`/`pod_name`/`container_name` labels and returned EMPTY / mismatched the old `banyandb::test-host` entities (31 failures). Regenerate banyandb-service/instance.data.yaml and add banyandb-endpoint.data.yaml with cluster-model mock input (cluster/pod_name/container_name/node_role/node_type + per-metric facets: kind/path/name/method/service/group/operation/type/le/seg) and expected entities (SERVICE=test-cluster, SERVICE_INSTANCE=test-pod@data, ENDPOINT=test-group). All 61 banyandb rules now pass MALExpressionExecutionTest locally. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…le in setup docs Review nits: (1) the e2e liaison scrape didn't set node_type, so labeled liaison metrics emitted node_type="" while the instance attribute is 'n/a' — inject node_type: n/a for the liaison target so the metric label matches the closure default. (2) Add the new banyandb-endpoint.yaml to the OpenTelemetry receiver setup table (was missing). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…proxy, not per-node scraping
Operator-facing gaps: the mandatory scrape job_name ('banyandb-monitoring', which every rule
file filters on) was undocumented, and the setup linked the no-FODC e2e config as the example.
Add an explicit FODC-proxy collector snippet (job_name + cluster injection + OTLP exporter), and
state plainly that operators scrape the FODC proxy — which resolves and stamps each node's
pod_name/container_name/node_role/node_type — rather than scraping nodes directly. Direct per-node
scraping (the e2e's setup, used only because it runs no FODC) is called out as not a production
pattern.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…B cluster case BanyanDB 0.11 (the bumped SW_BANYANDB_COMMIT) only serves queries within its retention window, so the shared cluster-cases.yaml's fixed 2022-01-26 downsampling timestamps return empty. Add a banyandb-only cluster-cases override that sends the meter data at recent timestamps and queries exactly one slot (start == end) per minute/hour/day, asserting the exact downsampled value for the slot the data was sent into. Cluster-up sanity sends are placed ~25h earlier (different day) so they never pollute the downsampling sums. Limited to test/e2e-v2/cases/cluster/zk/banyandb/ — the shared cluster-cases.yaml and cluster/expected/ are untouched, so the ZK ES case is unaffected. Verified locally: 12/12 pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
wankai123
approved these changes
Jun 11, 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
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.
Implement SWIP-15 BanyanDB self-observability around the cluster / container / group model
Rebuilds the
otel-rules/banyandb/self-observability feature to match how BanyanDB is actually built and operated (requires BanyanDB 0.11+):Service= one BanyanDB cluster (service(['cluster']));ServiceInstance= one container keyed onpod_name+container_name(withnode_role/node_type/container_name/pod_nameas instance attributes);Endpoint= one storage group.banyandb-endpoint.yaml;banyandb-service.yaml(7 rules) andbanyandb-instance.yaml(43 rules) redesigned. Source expressions mirror the upstream BanyanDB FODC-proxy Grafana boards (grafana-fodc-nodes.json/grafana-fodc-workload.json) so the SkyWalking dashboards stay in lockstep. The stale single-nodehost_namemodel and the removedetcd_operation_rate/up-derivedactive_instance/queue_sub_total_msg_sent_errmetrics are dropped.docs/en/banyandb/dashboards-banyandb.mdrewritten to the cluster/container/group model.:2121directly and injects the FODC-equivalent identity labels as static per-scrape-job labels, so the full rule set is exercised without a FODC proxy.Validation
DSLClassGeneratorTestcompiles every rule through the productionMetricConvertpath (service 7 + instance 43 + endpoint 11), including the 6-arginstance()properties closure,histogram_percentile,tagNotEqual, andtime().The e2e passes 16/16 against the latest BanyanDB 0.11 build, covering service / instance / endpoint scopes.
If this is non-trivial feature, paste the links/URLs to the design doc. — SWIP-15.
Update the documentation to include this new feature. —
docs/en/banyandb/dashboards-banyandb.md.Tests(including UT, IT, E2E) are added to verify the new feature. — rewritten
test/e2e-v2/cases/banyandb/e2e; rules compile-validated byDSLClassGeneratorTest.If it's UI related, attach the screenshots below. — N/A; the BanyanDB dashboards ship from the Horizon UI bundle (separate repo).
If this pull request closes/resolves/fixes an existing issue, replace the issue number. Closes #.
Update the `CHANGES` log.