Hi Willy, Could you please let me know whether the error with config check we're seeing in the 3.4.x branch is expected, or if a fix for it is also planned in the next patch release?
Thank you! On Jun 17, 2026 at 10:18:10, Willy Tarreau <[email protected]> wrote: > Hi, > > On Wed, Jun 17, 2026 at 02:06:49AM -0500, MailList Catch wrote: > > Hi, > > > I have issue with config check command: > > > # podman run --rm --name haproxy_cfg_test --user root -v > > /etc/haproxy:/etc/haproxy:ro docker.io/library/haproxy:3.4.0-alpine > haproxy > > -c -V -dD -dW -f /etc/haproxy/haproxy.cfg_backup > > [NOTICE] (1) : haproxy version is 3.4.0-64a335366 > > [DIAG] (1) : Created 8 threads split into 8 groups > > [ALERT] (1) : Some warnings were found and 'zero-warning' is set. > > Aborting. > > > End exit code = 1 > > > Minimal config: > > > defaults > > mode http > > timeout client 30m > > timeout connect 4s > > timeout server 30m > > timeout check 5s > > > frontend stats > > bind 127.9.0.1:9101 > > stats enable > > stats uri / > > > > On version 3.2.19 everything is ok: > > > # podman run --rm --name haproxy_cfg_test --user root -v > > /etc/haproxy:/etc/haproxy:ro docker.io/library/haproxy:3.2.19-alpine > > haproxy -c -V -dD -dW -f /etc/haproxy/haproxy.cfg_backup > > Configuration file is valid > > > Exit code = 0 > > > I can't understand where is problem with config check? > > > The issue is related to the combination of -dD with -dW. -dD detects > diag-level warnings, which are not necessarily issues but just possibly > suspicious config elements that are worth analyzing (for example you > could have two servers with the same addr+port which might be caused > by a copy-paste error or be intentional). You shouldn't use -dD in your > regular config checks, they're only meant to be used for debugging. > > The issue above started to trigger by default in 3.3, or in 3.2 when > specifying a cpu-policy because it uses ha_diag_warning() to report > the number of threads that were created. I addressed this last week > with these two commits that were marked for backporting to 3.2 so that > cpu-policy no longer uses ha_diag_warning() but a new ha_diag_notice(): > > 7d63efa5f ("MINOR: errors: add ha_diag_notice() to report diag-level > notifications") > 960fa1c92 ("BUG/MINOR: cpu-topo: use ha_diag_notice() to report thread > creations") > > Willy >
