Christopher Faulet

@wtarreau, it is gentle ping :)

We finally decided to not backport the series to 2.9. It is a more a performance issue than a bug. Now the 3.0 was released, it is probably better to...

2.0 is EOL, I'm closing now.

2.0 is EOL, I'm closing

Indeed, there is a leak at deinit. It is not a big deal. But it seems there is other leak on deinit. Is there any WURFL guys in the place...

I'm also unable to reproduce the issue with quick tests. @donnyxray, if you have a way to easily reproduce the issue, is it possible to enable health-check traces, at least...

On this trace, there is no loop, at least from the health-checks point of view. Could you confirm you observed a connect storm when this trace was produced ?

AFAIK, nobody is working on this issue. But I don't really understand the need. Probably because I'm not a Prometheus expert. Adding the same labels on all metrics seems strange...

The idea was not to add another exporter but to add a dedicated name for an specific haproxy instance, exported in a metric. Instead of having it as a label...

unfortunately no. The service must be modify to do so.