Ricardo Nabinger Sanchez
Ricardo Nabinger Sanchez
> Any problem since the upgrade ? Yes. After an unknown amount of time, some threads started burning a lot of CPU. We notice them easily because the network load...
> An interesting way to check this would be to duplicate the resolvers section and make half the servers use one and the other half the other. If that reduces...
The server is not too loaded, and the older HAProxy thread group is less active, but in spite of that, `resolv_process_responses` is taking measurably less effort. I've collected instructions instead...
Thanks, we will schedule a roll-out on our side.
> Any news about this issue ? I can't say for sure. This issue in particular has not popped up again, although we keep the mitigation on many of our...
Doing some rollbacks: * HAProxy version 2.9.7-a67b18-14 2024/04/08: broken * HAProxy version 2.8.9 2024/04/05: *WORKS!*
> Is there any good reason to delete the content-length header ? I'm doing this in order to stay on HTTP/1.0 + `Connection: close`, a common case we see. I...
When interfacing with Lighttpd, I have to manually remove, in order to get closer to our frequent use-case (end users in HTTP/1.0 mode, with unknown content-length). The case in this...
In my testing, I actually may have a `content-length:` header because I cannot completely reproduce our production environment where we easily see HTTP/1.0 responses relying solely on `connection: close`. In...
> Well it is not so easy because a regex may be used to remove a header or it may be removed from lua for instance. So in the end,...