Add filtering on server status #160
Conversation
|
Interesting, would you mind rebasing this against the current master? |
Allow excluding metrics for server slots depending on their status. Using this you can avoid reporting metrics for unused server slots when using dynamic scaling in haproxy. E.g. you can use --haproxy.server-exclude-states='DOWN,MAINT,MAINT (resolution)' Signed-off-by: Magnus Hyllander <[email protected]>
|
Rebased on master |
|
quay.io is having problems today. |
There was a problem hiding this comment.
Thanks a lot! I think we can keep the original interface of the Exporter and NewExporter() and do the filtering before, as we already have similar functionality. This will also avoid the changes to the test file as the interface stays the same.
A test for the filter function would be nice as well.
| return 1 | ||
| case "DOWN", "DOWN 1/2", "NOLB", "MAINT": | ||
| default: //case "DOWN", "DOWN 1/2", "NOLB", "MAINT", "MAINT(via)", "MAINT(resolution)": |
There was a problem hiding this comment.
We already return 0 / down as a fallback. What's the motivation for this change instead of listing known down status explicitly?
There was a problem hiding this comment.
I wasn't sure what to do with this function, since I haven't found a complete list of possible server statuses. I considered extending the list with the statuses I knew about, but decided there was no point, especially since (as you say) the fallback was 0 / down anyway. I guess I could have done it in a nicer way, letting it fall through to the fallback instead.
| } | ||
|
|
||
| for field, metric := range serverMetrics { | ||
| if _, ok := selected[field]; ok { |
* [CHANGE] Switch logging to go-kit #171 * [CHANGE] Fix metric types #182 * [CHANGE] Fix unit of time metric #183 * [FEATURE] Add filtering on server status #160 * [ENHANCEMENT] Add compression and server selection metrics #154 * [ENHANCEMENT] Add client/server abort metrics #167 * [ENHANCEMENT] Add version info metric (when using UNIX sockets) #180 Note: This release fixes the metric types of counters and renames the following metrics: * `haproxy_exporter_csv_parse_failures` -> `haproxy_exporter_csv_parse_failures_total` * `haproxy_exporter_total_scrapes` -> `haproxy_exporter_scrapes_total` * `haproxy_server_check_duration_milliseconds` -> `haproxy_server_check_duration_seconds` Signed-off-by: Tobias Schmidt <[email protected]>
* [CHANGE] Switch logging to go-kit #171 * [CHANGE] Fix metric types #182 * [CHANGE] Fix unit of time metric #183 * [FEATURE] Add filtering on server status #160 * [ENHANCEMENT] Add compression and server selection metrics #154 * [ENHANCEMENT] Add client/server abort metrics #167 * [ENHANCEMENT] Add version info metric (when using UNIX sockets) #180 Note: This release fixes the metric types of counters and renames the following metrics: * `haproxy_exporter_csv_parse_failures` -> `haproxy_exporter_csv_parse_failures_total` * `haproxy_exporter_total_scrapes` -> `haproxy_exporter_scrapes_total` * `haproxy_server_check_duration_milliseconds` -> `haproxy_server_check_duration_seconds` Signed-off-by: Tobias Schmidt <[email protected]> Co-authored-by: Tobias Schmidt <[email protected]>
Allow excluding metrics for server slots depending on their status. Using
this you can avoid reporting metrics for unused server slots when using
dynamic scaling in haproxy.
E.g. you can use
--haproxy.server-exclude-states='DOWN,MAINT,MAINT (resolution)'