Depends on D1276
Diff Detail
Diff Detail
- Branch
- master
- Lint
Lint Skipped - Unit
Unit Tests Skipped - Build Status
Buildable 4747 Build 6325: arc lint + arc unit
Event Timeline
Comment Actions
Build has FAILED
Link to build: https://jenkins.softwareheritage.org/job/DOBJS/job/tox/28/
See console output for more information: https://jenkins.softwareheritage.org/job/DOBJS/job/tox/28/console
| swh/objstorage/api/server.py | ||
|---|---|---|
| 28 | await is not a function, is it? | |
Comment Actions
In production, there's a statsd instance deployed on each machine, that gets scraped every minute by prometheus.
The data can then be used in a grafana dashboard.
| swh/objstorage/api/server.py | ||
|---|---|---|
| 25–28 | Rather than having such a high number of different metrics, please use a single metric name and add the function name as a label: swh_objstorage_request_count and swh_objstorage_request_duration_seconds, both with an endpoint label. This way we don't need to update a potential dashboard when we add a new function. I think we'll also want to set up a few static labels to be able to distinguish several instances of the objstorage on the host. | |
| 54 | Accumulating counters for volumes usually end with <unit>_total (https://prometheus.io/docs/practices/naming/) swh_objstorage_add_bytes_total or swh_objstorage_in_bytes_total would probably make sense. | |
| 78 | Same remark: swh_objstorage_get_bytes_total or swh_objstorage_out_bytes_total | |
Comment Actions
nowhere, unless one have a running statsd server. In our case, we do have a promotheus-statsd server running. So the answer to this question is: in promotheus (in prod).
| swh/objstorage/api/server.py | ||
|---|---|---|
| 25–28 | On the server-side, yeah, that'll work fine. | |