Re:Deep inside the Linux kernel: a network latency spike
by r4um on Friday November 20, 2015 @01:51AM
(#10600010)
The collapsing and pruning of tcp buffers shows up in network statistics also, could have been the first hint.
Usually system defaults work just fine, unless you want to tune for a very specific type of workload.
[ Parent
| on HN ]
Re:Deep inside the Linux kernel: a network latency spike
by majke on Friday November 20, 2015 @04:25AM
(#10600390)
Looking at netstat makes sense as long as the numbers are low. It's hard to correlate one of 167 counters to the unpredictable latency event you are just debugging :)
$ netstat -s | grep -E '(prune|collap)'
4445504 packets pruned from receive queue because of socket buffer overrun
150606 packets pruned from receive queue
278333723 packets collapsed in receive queue due to low socket buffer