Roman Gershman
Roman Gershman
Opened https://github.com/romange/helio/pull/343/files though this is not the root cause.
Did you try loading a dataset with X GB into a datastore with tiering and maxmemory Y, where Y
lgtm but it could be nice if you could run it one or two configs yourself and provide an anecdotal info on how well it works.
Consider adding to stats the event of skipping stashing due to low disk space
Also, please add a PR description! 🙏🏼
Is it fixed?
I prefer we keep this in a hackathon branch for now - so you will have freedom to experiment with half baked code.
@fernandomacho can you please attach the INFO logs from master and the replicas?
Try, increasing `--replication_stream_timeout` to 5000
how many CPUs do you use for your master and replica? what's the exact memtier that you use and exact arguments for both master and replica?