write_graphite stops sending to any node when one becomes unavailable

[chrisburrows]

I have a configuration using write_graphite to send data to two graphite destinations for resilience. If one of these destinations is unavailable, eventually no data is sent to the other functioning node. Data continues to be sent for a few minutes, but then no new data is sent.

Configuration is using 2 nodes with TCP transport and collectd 5.5.

[pulecp]

I can confirm that with the version 5.8.1. A connection times out but it's never recovered.

Jul 30 15:24:25 machine.example.com collectd[6997]: write_graphite plugin: send to carbon.example.com:2003 (tcp) failed with status 110 (Connection timed out)