Aurélien D.

It's not useful to dump the dictionary to troubleshoot the connection between the 2 peers. DICT is used to dump specific stick table data such as `server_key` Here from your...

>EDIT: I might want to add that 3.1-dev12, as opposed to 3.0, generates tons of warnings telling me about Legacy mailers used by backend xy will not be supported anymore...

Caused by 68cf3959b345ec9d596d40b01eb416d2bef6522c Looks like false positive. Prior to this commit srv pointer was retrieved using `__objt_server(s->target)` under `assign_server_address()` function. Now `alloc_dst_address()` takes srv pointer as parameter which is retrieved...

I'm still seeing "Success=1" with minimal conf: ``` echo reload | nc -U /tmp/ma Success=1 -- [NOTICE] (1161777) : Initializing new worker (1161786) [NOTICE] (1161786) : haproxy version is 3.1-dev0-0aee24-2306...

When issuing the reload manually from cli it would be interesting to check how long the reload takes

Looking at client native implementation, I'm wondering what could happen if cli response exceeds 1024 chars: https://github.com/haproxytech/client-native/blob/7a32e86b3ff90338802279abd10cc43d6d407901/runtime/runtime_single_client.go#L122 It looks like we break from the reading loop as soon as a...

@PKizzle by default socat has a .5 second inactivity timeout set, so I suspect with 3.2.0 something is making the startup/reload slightly slower and socat is timing out while it...

>It seems like 3.2.0 is tiny bit slower with parsing the configuration There is too little difference to draw conclusions I think, the addition of some keyword parsers or even...

Saw the link provided by Tim as a response to haproxy 3.1 release announce on the mailing list Now that we have the possibility to give custom names to logformat...

Before haproxy crashes (triggered by watchdog), a lot of these warning seem to be emitted: ``` WARNING! thread 1 has stopped processing traffic for 211 milliseconds with 3 streams currently...