The CI jobs rsync some cached materials from a central instance integration-castor03.integration.eqiad.wmflabs and there might be several jobs fetching from it at the same time.
The rsync compression has show up to be a bottleneck (found by Tim T188375#5378612) and has been removed. However we now are bandwidth throttled and I am thus willing to repplay rsync compression.
However, the instance is on cloudvirt1002 which has slow CPUs. An issue I found this summer (T223971).
So I am wondering whether we could move that instance to a newer cloudvirt which has faster CPUs. From some benchmarks I did on T223971, some cloudvirt are at least twice faster so would benefit rsync --compressed.
Ideally we would need to keep the same instance IP if that it is at all possible. Else I got to prepare some CI configuration changes.
The operation must be done when CI is not too busy since it is rather disrupting. Ideally during European morning.