Usage of TCP connections grows without bound

[dmuth]

So I started getting EADDRNOTAVAIL errors while testing our node_riak today.

I did some investigation, and it seems that that after inserting about 14,000 keys at a rate of about 300 keys/sec, I get the EADDRNOTAVAIL errors, and it looks like I have an insane number of open TCP connections:

dmuth@localhost:~/cliq (production) $ netstat -nt |grep :8098 |wc -l 25965

So it would seem like the same TCP connections are not being reused, but instead are just hanging in the TIME_WAIT state. Given the high rate at which I am inserting keys, my usage of Riak overtakes the ability of the those TCP connections to end.

Is this an actual bug, or am I doing something wrong on my end?

Thanks,

-- Doug

[DTrejo]

Hey, not sure what's going on here, but @mranney may have some idea. Hopefully he has a moment to throw in on this.

Cheers, D

[wlaurance]

Is this still an issue?