Hi Jérôme, On Tue, 4 Aug 2026 11:45:59 -0400 Jérôme Charaoui <[email protected]> wrote:
> We have been using pasta for rootless Podman networking on GitLab > runners, and have found that when using the version currently packaged > for trixie, on very busy runners with many containers spawning and > despawning, the process will eventually starts leave unclosed TCP > sockets that accumulate until we're plagued with connection errors from > within containers. Thanks for the report. Did you already identify a set of patches that would fix this? This smells like https://bugs.passt.top/show_bug.cgi?id=179, which would be fixed by these commits: https://passt.top/passt/log/?qt=range&q=eb3babf7b4cf3793e0ed48030db958ace6546d7e..d2f7c21cfb949f2b1587b9475917efdd6ac549fd I haven't checked how involved a backport would be in this case, even though at a glance I'd say it should be doable. -- Stefano
