Am 15.05.25 um 10:41 schrieb Fiona Ebner: > Am 14.05.25 um 16:33 schrieb Thomas Lamprecht: >> Am 13.05.25 um 12:56 schrieb Fiona Ebner: >>> When the service is restarted, the directory will be recreated. The >>> issue is that the /run/pve directory is not just used as the runtime >>> directory of this service, but also for other things, e.g. storage >>> tunnel and mtunnel sockets and container stderr logs as well as pull >>> metrics, which will be lost when the service is restarted. >> Is there much gained by moving the socket path or would it be enough to >> drop the RundtimeDirectory completely or do we want to rely on the >> clean-up being done by systemd? Which probably is an OK enough reason, >> I'd think.. > > Yes, the cleanup aspect is there and it seems nicer to have a dedicated > directory IMHO. It would avoid the need to adapt pve-container and the > postinst snippet, but I'd still like to go for the "dedicated > RuntimeDirectory" approach.
Alright with me. >> btw., to name all options, there would be the RuntimeDirectoryPreserve= >> property for systemd services (documented in man systemd.exec) which >> we could set to `restart` to handle just updates or `yes` to never clean >> it up; besides obviously a system reboot clearing any tmpfs, that is. > > That option would not help with the awkward dependency/reliance on the > service to create the /run/pve directory in the first place. And even if > it's created by the tmpfiles.d snippet in pve-manager first, it will get > overwritten (with root:root 0755 permissions) when its first used as the > service's runtime directory. Fair enough, then let's go with your proposed approach. _______________________________________________ pve-devel mailing list pve-devel@lists.proxmox.com https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
