Seeing workers use 3-8GB of ram which is exceedingly high, I normally see them a couple hundred MB or less.
This has never happened before so I can’t replicate it just yet.
It’s affecting my consumer servers mostly, which those servers will take a produced tracks over PipeTransport and consume it for many viewers. I don’t see this problem on the producer server which suggests that consuming and producing aren’t the true cause for this ramp-up potentially but may be (consumer servers can open 1-1,000+ transports in a matter of seconds and clear immediately depending on room size and people entering/leaving rooms)
Would be curious if this has anything to do with PipeTransport and retransmission. Last week and a half servers had just started having major latency issues; 90ms average but every 10-30 seconds a 1,000-3,000ms increase. Thinking this may be slowly adding up over-time or something isn’t clearing from memory.
Open to ideas to test this locally and report findings.