| *SOLVED* Out of Socket Memory on the Hardware Node [message #10220] |
Mon, 12 February 2007 22:24  |
jubo
Messages: 8
Registered: December 2006
|
Junior Member |
|
|
I am running about 10 VEs on a hardware node. These are primarily web servers that get a decent amount of traffic. When I check the /proc/user_beancounters I notice that the following values sometimes go over limit: tcpsndbuf,tcprcvbuf,othersockbuf. So naturally I raise them up. However, now I am starting to get some Out of Socket Memory Errors in the logs on the Hardware Node. I did the following:
Added these increased values to the sysctl.conf
net.core.rmem_max=262144
net.core.wmem_max=262144
However, I still get the error, so I was wondering if people could suggest the correct value to use here. Is the sum of all the buffer sizes of all the VEs on the hardware node the correct value?
Also, what are safe values to use if you want to ensure little or no loss of service on a web server VE if it encounters a lot of traffic.
Thanks
[Updated on: Thu, 15 February 2007 10:15] by Moderator Report message to a moderator |
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
