OpenVZ Forum


Home » General » Support » TCP: too many of orphaned sockets (1 in CT102) - Debian
TCP: too many of orphaned sockets (1 in CT102) - Debian [message #46049] Tue, 24 April 2012 12:33
helmut666 is currently offline  helmut666
Messages: 1
Registered: April 2012
Location: Poland
Junior Member
Hi guys,
after deploying in production our new server with OpenVZ, it turns out that Linux kernel floods on console :

[cut]
[36898.990806] TCP: too many of orphaned sockets (1 in CT102)
[36899.045800] TCP: too many of orphaned sockets (1 in CT102)
[36899.272960] TCP: too many of orphaned sockets (1 in CT102)
[36899.290431] TCP: too many of orphaned sockets (1 in CT102)
[36903.551981] __ratelimit: 37 callbacks suppressed
[cut]
Some technical details:
Debian squeeze (6.0.4) with kernel 2.6.32-5-openvz-amd64
Hardware: Dell PowerEdge R910, 128GB RAM.
Kernel starts flooding about 10-15 min after turning on the container. I'm afraid because kernel realy floods with this messages (it's not occasional warning).
On the other hand - all services in this container are working properly and fast. There is no failcnt in /proc/user_beancounters.

So, i have 2 questions connected with this:
1. Should we be worried about it?
2. How about general system performance? (how intense kernel flooding influence on system performance). Is it possibile disabling this flooding?

Below is my container's config file:

KMEMSIZE="400000000:480000000"
#LOCKEDPAGES="512:512"
LOCKEDPAGES="999999:999999"
PRIVVMPAGES="1310720:1835008"
VMGUARPAGES="786432:9223372036854775807"
NUMPROC="600:700"
PHYSPAGES="0:9223372036854775807"
OOMGUARPAGES="786432:9223372036854775807"
NUMTCPSOCK="880:1000"
NUMFLOCK="188:206"
NUMPTY="16:16"
NUMSIGINFO="256:256"
TCPSNDBUF="107374592:289400832"
TCPRCVBUF="107374592:289400832"
OTHERSOCKBUF="1126080:2097152"

I cannot answer if other containers have the same problem, because for now this is the only one (i suppose, with other containers the problem appears too).
Inside this container works Apache with PHP and Postfix (the load inside container is rather light)

thanks a lot for suggestion. maybe it's a bug?
Previous Topic: allow firewall to pass all traffic for containers?
Next Topic: OpenVZ and rpc.idmapd (NFSv4)
Goto Forum:
  


Current Time: Sun Jul 20 19:31:39 GMT 2025

Total time taken to generate the page: 0.06486 seconds