Aurélien D.

>From my understanding, [BUG/MEDIUM: server/cli: don't delete a dynamic server that has streams](https://git.haproxy.org/?p=haproxy-2.8.git;a=commitdiff;h=4bc0909) will solve the crash, however, it will fail to delete the server and it will output Server...

>shutdown sessions server OLD_SERVER Unfortunately I think it will do as if `on-marked-down shutdown-session` was set and server is put under maintenance: `srv_shutdown_streams()` will be called, which will cause `stream_shutdown()`...

Regarding the potential leak, in practice it doesn't seem to happen because default proxies get referenced in proxies that inherited from them. And when those child proxies are freed, there...

Thanks for the info! Then it might be somewhat related to this recent commit: https://github.com/haproxy/haproxy/commit/cf07cb96be70cbe979c6cf21aefd134894cf743f Although I don't see anything wrong with the commit.. So perhaps coverity lost the track...

Thanks for the report It would be nice to see if you still encounter the issue in 2.7.8: ~150 bugs were addressed since 2.7.4 (http://www.haproxy.org/bugs/bugs-2.7.4.html), including bugs that refer to...

Great, thank you for the feedback. > But let's wait for a day or two before celebrate (and close the issue) Yeah sure, no problem!

Were the perf top and the various commands (show profiling tasks, activity..) taken during the surge (when haproxy is stuck at 100%)? Also it seems that you used `show profiling...

May I suggest another test @ut0mt8? When you notice an instance going rogue (haproxy @100%), do you think you would be able to temporarily redispatch the load (like redirect or...

Indeed, table cli actions (show, set, clear) were probably not taken into account when support for stick table array types was added in c64a2a307ccd304f0e5ee5a842b019cf6d4053b1. With the current implementation only the...

Seems related to https://github.com/haproxy/haproxy/issues/1886