rk@tigase.net opened 2 days ago
-
Approach I plan to use add Prometheus to instrument and display metrics from various parts of Sztab.
I am hence adding as a part of this issue: add
programming language prometheus/client_golangto Sztabina to Expose /metrics endpoint on :8085 Instrument forwardWithRetry (notify pipeline) Instrument GitHTTPHandler (clone/push/fetch duration) Add Sztabina as second scrape target in prometheus.ymlComplication notify runs as an ephemeral subprocess — counters in subprocess memory die with the process and never reach /metrics. Fix: notify subprocess POSTs a NotifyEvent to main process /internal/metrics/notify before exiting.
-
Tasks
- Add prometheus/client_golang dependency
- metrics/metrics.go — define counters and histograms
- git_http_handler.go — instrument git HTTP requests
- main.go — register metrics, expose /metrics, add /internal/metrics/notify handler
- notify.go — POST NotifyEvent to main process instead of direct counter increment
- Verify sztabina_notify_attempts_total appears in Prometheus after a push
-
Previous Value Current Value Open
In Progress
-
-
Sztabina instrumented with prometheus/client_golang. notify pipeline metrics (attempts/success/failure) working via subprocess=>main process POST relay. Git HTTP handler instrumented for request count and duration. Both targets UP in Prometheus. Closing with 1.10.1.
-
Previous Value Current Value In Progress
Closed
| Type |
New Feature
|
| Priority |
Normal
|
| Assignee | |
| Version |
1.10.1
|
| Sprints |
n/a
|
| Customer |
n/a
|
Problem The pressure of Sztabina into Sztabina is a critical junction and can break when dealing with large repos. Sztabina has no metrics exposure. To characterize the Sztabina => Sztab boundary under large repo load — and justify streaming — we need to measure notify attempt rate, success/failure, and git HTTP operation duration.