These are chat archives for atomix/atomix

23rd
Oct 2018
Jordan Halterman
@kuujo
Oct 23 2018 00:26
@echavez the logs always replay yes. That’s how the state is restored. The new Raft implementation replays fewer logs than the old one, which retained more log entries and stored fewer/smaller snapshots. The time it takes to restore a node’s state just depends on how many log entries it has to replay, but it shouldn’t take that long. As long as the nodes are not overloaded, the Raft logs should stay around 64MiB by default.
furtunately we have tests for this sort of thing, and I’m running them now
Jordan Halterman
@kuujo
Oct 23 2018 00:32
running a test that bootstraps a cluster with 100k writes and then kills and restarts nodes
need to make it possible to mess with log compaction in the tests too
to force it to retain the logs
and compact them as specific times
I think I have a patch for that around here somewhere...
Alex Robin
@alexrobin
Oct 23 2018 08:55
@kuujo did you have time to look at my updates of PR #863?