2026-10-01 –, Loft
When a machine/server/node is serviced and experiences a downtime, that results in a service interruption, customers or users tend to get quite sad. This talk will explore two recent additions to systemd that aim to make these service interruptions as minimal as possible.
systemd v260 and v261 added two features which aim to let service owners minimize downtime.
Firstly, when services use dm-verity images, profiling data showed that the largest bottleneck by far was loading the new images, which used to happen after the running service had already stopped during a unit restart. Since v260, when very specific conditions are met, this bottleneck has been removed from the critical path.
Secondly, a common issue causing slowdowns for services is having to serialize and deserialize large amount of state from slow disks after a reboot due to an OS update. Since v261, systemd supports the kernel's Live Update Orchestration, and uses it to persist the content of the FD Store across a kexec reboot. Services can use a memfd for their runtime state, hand it over the systemd when they are stopped/restarted, and they will get it back and can mmap it and use it straight away on restart. With LUO support, services can rely on this even across a kexec, removing the need to use serialize and deserialize state to disk.
This talk will explore both features, explaining how they work and how they can be used.
Software engineer at Microsoft by day, open source developer involved in various projects by night (systemd maintainer, DPDK LTS maintainer, ZeroMQ project co-lead, ...)