Konstantin Shcherban

@grobian i've tried increasing expire time to 90 and it didn't help. Do you need some additional information?

Collectd is sometimes sending incomplete data, like this. But the overall amount of correct data is greatly higher and if i manually render sum in graphite web it looks fine,...

Ignoring NaN makes sense as it's garbage and aggregated metric shouldn't be affected by some seldom garbage. Strange that collectd sends garbage over TCP, looks like bug.

This is what i've got from tcpdump. Log is not complete and i changed metric names ``` tcpdump -A -i eth0 port 2013 | grep 'collectd.servers.*if_octets' collectd.servers.x.x.x.x.interface-eth0.if_octets.rx 489.966782 1472125946 collectd.servers.x.x.x.x.interface-eth0.if_octets.tx...

nope, i use init script from debian package and it has only restart option

Yep, i still see nan from aggregator.

It's happening not too often. You can check this [affected graph](https://www.prolinux.org/files/c-relay-aggregator.png).

Thanks for reporting, i will have a look. Tested only locally with environment variables.

@scottharman Can't you install it from pip there? What's the error? Have you tried [this ](https://stackoverflow.com/a/11405769)solution? Another possible fix: https://github.com/pycrypto/pycrypto/issues/218#issuecomment-581757162

Struggling without `ssh` build feature in docker-compose.