Following up on earlier notes, I decided to make one more change: I doubled the sending buffer’s size.
[/usr/home/paul]:: sysctl net.inet.tcp.sendspace
net.inet.tcp.sendspace: 131070
I rebooted the system that’s been taking all the load and these tuning adjustments weren’t permanent at that point. It took a very short time before I saw the same problem cropping up. I reset the send and receive buffers to 64k and things improved. I don’t know if I need a send buffer that large, but I don’t think it will hurt anything. I do have some transaction in today’s log of more than 200k, one of more than 2 Mb.
Now playing: The National Anthem by Radiohead from the album “Kid A” | Get it