unit-e icon indicating copy to clipboard operation
unit-e copied to clipboard

Unit-e stats unable to open socket to connect to peer

Open Ruteri opened this issue 6 years ago • 0 comments

Describe the bug After 1.5h of running united in testnet I started getting a lot of logs, and the node does not connect to it's peers (it was connected previous to the errors), and the node it tries to connect to does not experience any problems.

Logs I got (roughly 5 per second): 2019-01-25 15:13:08 [ ] Cannot create connection: non-selectable socket created (fd >= FD_SETSIZE ?) 2019-01-25 15:13:08 [ net] trying connection 35.231.131.63:17182 lastseen=0.0hrs

Running lsof | grep united on the host outputs 3.5k lines of (mostly) UDP sockets: united 5755 5847 root 1141u sock 0,9 0t0 45889 protocol: UDP Running lsof inside the container results in over 1k sockets: 1 /opt/unit-e/bin/united socket:[45862] The sockets do not show up in netstat.

To Reproduce It only happened once on one of the nodes:

  1. Run node in testnet
  2. Wait

Expected behavior Node does not stop connecting to peers.

Environment

  • OS: Ubuntu 18.04
  • UnitE version: checked out on 44431c3d03fb12b5651aa30d4ec8976b34d0e9b6
  • Runs in docker on gce server

Additional context The problem is most likely caused by not cleaning up the sockets. I did not kill the node, so I can provide more info.

Edit: added lsof from inside the container.

Ruteri avatar Jan 28 '19 10:01 Ruteri