Darn, you beat me to it ^^ That does feel like a good improvement!
For reference, here is a little PutObject benchmark for this change.
Before (as seen in #851) :
![image](/attachments/ff…
Thank you very much for your feedback!
I'm just worried that with preemptive send enabled, in the case of slow machines with fast networking, the watchdog would trigger nearly all of the…
Quick follow-up on the previous benchmark, discussion below - thanks again everyone for your input!
Inlined objects, no data logic
Try again with objects at 1KB to remove the…
Retest on a cluster w/ data blocks (objects beyond 3KiB)
Very nice catch for the read ahead flag, thank you for your commit! To further investigate the issue, we re-ran a benchmark of our…
Hello,
Definitely with you regarding metrics, I've just pushed a commit adding rpc_watchdogs_started_counter
and rpc_watchdogs_preemption_counter
. I've adjusted the names slightly in hope…