Valentín Gutiérrez

@wtarreau thanks! I'll test haproxy 2.8 on Monday and report back

running HAProxy 2.8.5 now: ``` HAProxy version 2.8.5-1~bpo11+1 2023/12/09 - https://haproxy.org/ Status: long-term supported branch - will stop receiving fixes around Q2 2028. Known bugs: http://www.haproxy.org/bugs/bugs-2.8.5.html Running on: Linux 5.10.0-25-amd64...

@wtarreau that's a super easy test given that the increased CPU load is quite steady lately:  the impacted node is cp3066.. as soon as I depooled it today to...

> OK thanks very much! If it moves to another node, it could be related to always the same source address. I guess you're having hash-based L3 or L4 load...

we've detected the same issue with mtail 3.0.0-rc50 (shipped with Debian bookworm) and latest version available (3.0.0-rc55)

perf top report looks like this: ``` 14.22% mtail mtail [.] regexp.(*Regexp).tryBacktrack 8.83% mtail [kernel.kallsyms] [k] finish_task_switch.isra.0 3.73% mtail mtail [.] runtime.mallocgc 3.58% mtail [kernel.kallsyms] [k] _raw_spin_unlock_irqrestore 2.79% mtail mtail...

> > A thread got stuck for no less than 30 seconds while trying to free memory, without even receiving signals for all this time! It makes me think that...