| Container networking breaks after a while [message #42970] |
Thu, 23 June 2011 08:57  |
DavidKnell
Messages: 2
Registered: June 2011
Location: Greece
|
Junior Member |
|
|
I'm running a number of containers for VoIP traffic, and I'm having a certain amount of trouble with those which are running FreeSWITCH. Interestingly, one running Asterisk with about the same amount of traffic is working fine.
What happens, after a while, is that the container's networking just stops. Name resolution ceases; attempts to ping the outside world (by IP or hostname) fail, tshark within the container shows no network traffic in or out, tshark on the host node's Ethernet interface shows packets inbound to the container, but nothing emanating from it.
The beancounters show overruns on tcpsndbuf, tcprcvbuf and dgramrcvbuf.
Restarting the container fixes the problem for a while - typically about 24 hours.
The host's running x86_64 kernel 2.6.18-238.9.1.el5.028stab089.1
Any ideas, pointers, or things to take a look at greatly appreciated.
--Dave
|
|
|
|
| Re: Container networking breaks after a while [message #42974 is a reply to message #42970] |
Thu, 23 June 2011 11:17  |
DavidKnell
Messages: 2
Registered: June 2011
Location: Greece
|
Junior Member |
|
|
An update.
The above isn't strictly correct. A failed container:
- responds to PINGs
- will respond to an inbound SSH connection by opening the TCP connection OK, but won't then go on to try to establish anything above the basic TCP session
- doesn't show anything at all within the container using wireshark
So it looks a bit like there's something hung up half-way up the network stack. netstat shows a bunch of sockets with stuff in their receive queues; these amounts are constant and are sufficient to fill up the container's dgramrcvbuf.
Here's the final weirdness: killing FreeSWITCH within the container fixes things.
--Dave
[Updated on: Thu, 23 June 2011 11:45] Report message to a moderator |
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
