Storage replication concurrency

[Rens]

I couldn't find any means of managing the amount of concurrent replication jobs running from a host. Of course it's important to honour the schedule as close as possible, but if you have 100 VM's on one host you want to replicate all of them (could of course use low level pool replication) would you manually have to spread them over a day, to prevent too many concurrency? As far as i can see, bandwidth limiting is per host so concurrent bandwidth and remote write IO could still be high.

Just wondering

 

[brwainer]

Assuming it is based on the guts of pve-zsync, which is appears to be, then it should have the same implementation of only having a single job per source-host at a time, with jobs being tacked onto a queue as their scheduled time passes. But this needs confirmation.

 

[wolfgang]

Only one job per host is replicating at the same time.

 

[Rens]

That's exactly what i needed to hear tnx...