Lanqing Yang
Lanqing Yang
After reading the [doc](https://www.datadoghq.com/blog/go-memory-metrics/ ) i am fairly confident on all the new metrics from runtime/metrics. This is a risky change at the end of the day, Hi @nvanbenschoten do...
yes exactly I want to use a billing mode calculate how many tokens after we ran the requests.
was about to achieve great snapshot throughput with adjusted snapshot concurrency on a 9-9-2 cluster (replication 3-1-1) by docker pause on a node on chi and turn it back on...
re-reading nathan's comment maybe the low snapshot generation rate was due to the delegated snapshot, let me try pause a node in ash1, and set lease holder preferences to none...
new testing results with delegated snapshot and lease preference disabled one node received up to 1.5 gb of snapshot data per second, the 7hr paused node was up to speed...
> @lyang24 how did the recipient of the 1.5GB/s of snapshots fare? Was it overloaded? Did its LSM retain its shape or did we see read amplification? I don't see...
Just chatted with Roblox colleagues this is quite helpful on our cockroachdb migration process which migrated data to 3 copies in ASH and then changes the leaseholder to distribute data...
spoke with @andrewbaptist I will address the follow ups in later prs. bors r+