arcadedb / helm
CLAUDE.md1 starsChanged 5 months agoApache-2.0View on GitHub
Directory rules for the ArcadeDB cell chart. Inherits the root CLAUDE.md. Validate with helm lint + kubeconform; never helm install against a real cluster in Phase D. - Image: image.tag must be a semver ≥ 26.4.1, never latest; prefer pinning image.digest in prod (the chart fails on a violation). - Replicas: 3 in prod (PDB minAvailable: 2). Non-prod may be 1. - One pod per AZ: hard podAntiAffinity on topology.kubernetes.io/zone; schedule onto the DB nodes via nodeSelector: {workload: arcadedb} +…
In this repository20 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- control-plane/CLAUDE.mdCLAUDE.md
- docs/CLAUDE.mdCLAUDE.md
- terraform/CLAUDE.mdCLAUDE.md
- add-cellSkill
- cell-capacity-reportSkill
- deprovision-tenantSkill
- dr-drillSkill
- incident-triageSkill
- migrate-schemaSkill
- new-adrSkill
- provision-tenantSkill
- restore-tenantSkill
- retire-cellSkill
- review-terraform-planSkill
- rotate-secretsSkill
- security-baseline-checkSkill
- tenant-erasureSkill
- tenant-usage-reportSkill
- upgrade-arcadedbSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 48
Do not rely on the bare StatefulSet RollingUpdate.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
CLAUDE.md — helm/
Directory rules for the ArcadeDB cell chart. Inherits the root
CLAUDE.md. Validate with helm lint + kubeconform; never
helm install against a real cluster in Phase D.
ArcadeDB values conventions
- Image:
image.tagmust be a semver ≥ 26.4.1, neverlatest; prefer pinningimage.digestin prod (the chartfails on a violation). - Replicas: 3 in prod (PDB
minAvailable: 2). Non-prod may be 1. - One pod per AZ: hard
podAntiAffinityontopology.kubernetes.io/zone; schedule onto the DB nodes vianodeSelector: {workload: arcadedb}+ the matching toleration.
The sizing rule (THE #1 gotcha — prime directive #7)
sizing.podMemoryLimitGib must be ≥ maxPageRAMGib + heapGib + overheadGib.
The chart refuses to render otherwise (template fail). Give most RAM to the
off-heap page cache (maxPageRAM), keep heap modest, and leave cpuLimit
empty — a tight CPU limit throttles the JVM and starves Raft heartbeats →
leader flapping.
Worked example (r7g.2xlarge, 64 GiB): maxPageRAMGib: 32, heapGib: 8,
overheadGib: 6, podMemoryLimitGib: 46.
txWalFlush tiering (ADR-0013)
arcadedb.txWalFlush: 2 (fsync) for enterprise/regulated, 0/1 for
standard (throughput). Set per cell; do not default-and-forget.
Probes
/ready (HTTP 204, no auth) for readiness/liveness; generous startup probe
(failureThreshold high) — ArcadeDB + index load can be slow on big DBs.
The /prometheus MIME-type workaround (F6 / assumption A17)
ArcadeDB returns Content-Type: application/json for /prometheus. Use ONE of:
- Scraper-side (preferred): force the text parser
(
fallbackScrapeProtocol: PrometheusText0.0.4in the PodMonitor, or the ADOTfallback_scrape_protocol). - Sidecar:
metrics.mimeWorkaround.sidecar.enabled=true. Revisit at every version bump — drop it if a release fixes the bug.
The quorum-preserving upgrade (no Operator → we own it — ADR-0029)
Do not rely on the bare StatefulSet RollingUpdate. Upgrades run via the
quorum-aware Argo workflow (upgrade-arcadedb skill): pre-flight health + fresh
backup → replicas first, one at a time (wait for re-join + lag→0 + /ready)
→ leader last (graceful step-down) → re-audit cross-DB isolation.
Forward-only: rollback = restore-from-backup to the prior version.
Other
- Root password is set-once (F4): injected from a secret at first boot only;
"rotate root" = provision a new admin user (
rotate-secretsskill). - Never a public LoadBalancer — the client Service is ClusterIP; external app access is via the platform API → internal NLB → PrivateLink (ADR-0026).
- NetworkPolicies are owned by the Terraform
cellmodule; don't duplicate here.
Validate
helm lint helm/arcadedb
helm template arcadedb helm/arcadedb | kubeconform -strict -ignore-missing-schemas -summaryDoes it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

