OpenVZ Forum


Home » General » Support » TCP: time wait bucket table overflow - memory leak?
Re: TCP: time wait bucket table overflow - memory leak? [message #36752 is a reply to message #36745] Wed, 15 July 2009 14:36 Go to previous messageGo to previous message
nksupport is currently offline  nksupport
Messages: 16
Registered: June 2007
Junior Member
maratrus wrote

But there is no such a killer in OpenVZ. Parallels Virtuozzo Containers contains subsystem which does what you have said but in OpenVZ you can adjust the behavior of the container via user_beancounters i.e. if barriers and limits are really huge there is no a restrictive force which prevents a containers from consuming a lot of resources.



Yes there is! A process that failed to allocate memory will onviously be killed, isn't that right? I am talking exactly about UBC in my post. I'm using UBC to limit the VEs. Like i said, the current UBC limits are drakonian. I want all processes that try to overuse memory to be killed and OpenVZ is supposed to do it.


maratrus wrote

Could you please say what means "the entire node just dies"? Are you able to ping the node? Are you able to invoke commands?
What is the reason of LA being so big?
Does the node perform CPU-consuming operations? What is the CPU state when "the node is being died"?
Or there are a lot of input-output operations being invoked?



You know, it's hard to debug a server at LA of 500 Smile I can still ping it, but it's just too slow. Random processes from VEs are in top, it behaves just like it would with unlimited UBC, i.e. instead of killing or throttling the runaway process it gives it more and more CPU. Eventually the node's ssh dies and this ends the show. Anyway - i attached my /proc/user_beanconters and performance stats, please have a look if i missed anything.


"It's the power cord", I say
 
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message icon5.gif
Read Message
Read Message
Previous Topic: number of cpus on a x64 kernel
Next Topic: dump broken in Debian 5 lenny?
Goto Forum:
  


Current Time: Sun Jul 06 23:57:36 GMT 2025

Total time taken to generate the page: 0.03036 seconds