Skip to content

Self-hosted: namespace selection, runtime tag fix, non-k8s metrics bridge - #3

Merged
technoidentity-kuncha merged 1 commit into
technoidentity:mainfrom
PardhuTI:feat/self-hosted-namespace-and-bridge
Aug 24, 2026
Merged

Self-hosted: namespace selection, runtime tag fix, non-k8s metrics bridge#3
technoidentity-kuncha merged 1 commit into
technoidentity:mainfrom
PardhuTI:feat/self-hosted-namespace-and-bridge

Conversation

@PardhuTI

Copy link
Copy Markdown
Contributor

Follow-up to #2, all validated against a live Temporal Server 1.31.2 deployment (metrics scraped from the server /metrics endpoint and confirmed in Sumo).

Changes

  • Namespace selection: namespace is now a multi-select dropdown (populated from metric=persistence_requests). Namespace filtering is applied to every namespace-bearing panel (service, workflow/task, matching/backlog, persistence). Cache and runtime panels are left global because those metrics carry no namespace label — filtering them would blank the panels.
  • Runtime tag fix: restarts, num_goroutines, memory_heap are grouped by service_name — the tag the server actually emits — not temporal_service_type (which Temporal's published Grafana dashboard used). Docs synced.
  • Non-Kubernetes metrics bridge: sumo/temporal-self-hosted-metrics-bridge.sh + temporal-self-hosted-metrics-bridge.service forward the server /metrics to a Sumo HTTP Logs & Metrics source for EC2/VM/Docker deployments (gzip pull to avoid WAN truncation; only dashboard families forwarded, _bucket dropped to bound cardinality). Kubernetes deployments keep using the annotation scrape.

Validation

Live server confirms bare (Tally) counter names — no _total; all dashboard families present with data across namespaces. In Sumo, namespace-scoped queries return data (e.g. service_requests namespace=ameriprise_agentic). No changes to the Cloud or Worker dashboards.

…ics bridge

Validated against a live Temporal Server 1.31.2 deployment:
- Add a multi-select Temporal Namespace variable and apply namespace filtering to
  every namespace-bearing panel (service, workflow/task, matching/backlog,
  persistence); leave cache/runtime panels global since they carry no namespace.
- Group server runtime metrics (restarts, num_goroutines, memory_heap) by
  service_name, the tag actually emitted (not temporal_service_type); sync docs.
- Add temporal-self-hosted-metrics-bridge.sh + systemd unit to forward server
  /metrics to a Sumo HTTP source for EC2/VM/Docker (non-Kubernetes) deployments.
@technoidentity-kuncha
technoidentity-kuncha merged commit ba95782 into technoidentity:main Aug 24, 2026
1 check failed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants