OpenVZ Forum: Support » High number of /proc/user

Home » General » Support » High number of /proc/user_beancounters failcnt

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

High number of /proc/user_beancounters failcnt [message #27462]

Mon, 18 February 2008 09:56

ittec
Messages: 41
Registered: February 2008

Member

Hi again

i read a lot of threads about OpenVz and its memory sistem and how you can see the guaranted memory limit for each VE. I know that, privvmpages assign the total of "real" memory to VE as:

TotVeRam = privvmpages * 4

So if you have

privvmpages = 155119, the total of RAM to one VE is 620476 Kb, as you can see into VE when you run a top or you make #cat /proc/meminfo (into Ve)Well, this number of privvmpages is a result of a split(5) one HN but this VE have a lot of traffic and # cat /proc/user_beancounters show the next:

privvmpages

privvmpages 33870 171049 155119 170630 647449

As i can read in another post, privvmpages:

1. provide any guarantee for memory allocation
2. provide a strong limit for the memory consumption of your VE

My question is, based in number of privvmpages fails how i setup the privvmpages of VE conf to guarantee more memory allocation? I ask this because I did a --split 5 of HN so if now i modify this value only to one VE im breaking this split. At the moment I try to increase this number with caution but i have no reference about how do it.

Thanks Smile

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27466 is a reply to message #27462]

Mon, 18 February 2008 10:13

maratrus
Messages: 1495
Registered: August 2007
Location: Moscow

Senior Member

Hi,

sorry, but can you please specify what does it mean

Quote:

I did a --split 5 of HN

?
You might have used vzsplit utility.

You can increase barrier and limit for privvmpages with vzctl utility.

vzctl set  <VE_ID>  --privvmpages <VAL1>:<VAL2>

VE_ID - your VE ID
VAL1 - barrier value
VAL2 - limit value

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27468 is a reply to message #27466]

Mon, 18 February 2008 10:28

ittec
Messages: 41
Registered: February 2008

Member

Quote:

Hi,

sorry, but can you please specify what does it mean
Quote:

I did a --split 5 of HN
?

Hi maratrus

I created the VE with this command:

vzctl create 249034 --ostemplate centos-4-i386-minimal --config split5

I know how increase the barrier and limit of privvmpages but Im confused about which values I need to setup to avoid failcnt troubles. Increasingly the barrier and limit of VE to assign about 256MB more each time? Im not sure if Im explaining me good excuse me.

Thanks.

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27472 is a reply to message #27468]

Mon, 18 February 2008 10:57

kir
Messages: 1645
Registered: August 2005
Location: Moscow, Russia

Senior Member

See http://wiki.openvz.org/Resource_shortage

Kir Kolyshkin

http://static.openvz.org/userbars/openvz-developer.png

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27474 is a reply to message #27468]

Mon, 18 February 2008 11:14

maratrus
Messages: 1495
Registered: August 2007
Location: Moscow

Senior Member

Hi,

I can suggest you such decision.
Set barrier and limit to huge values. Start your applications inside VE and then look at UBC via /proc (held and maxheld columns). In such way you can obtain approximate consumption of privvpages. Then you can decrease privvpages.

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27479 is a reply to message #27472]

Mon, 18 February 2008 12:18

JimL
Messages: 116
Registered: February 2007

Senior Member

kir wrote on Mon, 18 February 2008 05:57

See http://wiki.openvz.org/Resource_shortage

Hi, Kir,

I think what he is trying to ask is more along the lines of "how big should I make it and if I guess too big, what will be the consequences?" I think Matarus is on the right track, but the question is still "what are the consequences of running a resource too high?" Especially if you have a number of VEs running into limits.

The reason I jumped in is that I've always been concerned about this issue too.

Jim.

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27487 is a reply to message #27479]

Mon, 18 February 2008 13:38

kir
Messages: 1645
Registered: August 2005
Location: Moscow, Russia

Senior Member

JimL wrote on Mon, 18 February 2008 15:18

I think what he is trying to ask is more along the lines of "how big should I make it and if I guess too big, what will be the consequences?" I think Matarus is on the right track, but the question is still "what are the consequences of running a resource too high?" Especially if you have a number of VEs running into limits.

The reason I jumped in is that I've always been concerned about this issue too.

Right.

System-wide UBC constraints are explained in great details at http://wiki.openvz.org/UBC_systemwide_configuration. Also, there is a utility that shows you some numbers using the formulae from the above article.

Kir Kolyshkin

[Updated on: Mon, 18 February 2008 14:40]

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27488 is a reply to message #27487]

Mon, 18 February 2008 14:20

JimL
Messages: 116
Registered: February 2007

Senior Member

Hi, Kir,
Thanks.

Quote:

Note the url Kir included has an extra period which leads you to nowhere.

It should be: http://wiki.openvz.org/UBC_systemwide_configuration

Jim.

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27490 is a reply to message #27488]

Mon, 18 February 2008 14:42

kir
Messages: 1645
Registered: August 2005
Location: Moscow, Russia

Senior Member

Thanks for correction. I have fixed the URL in my original comment.

Kir Kolyshkin

Report message to a moderator

Re: High number of /proc/user_beancounters failcnt [message #27495 is a reply to message #27474]

Mon, 18 February 2008 15:58

ittec
Messages: 41
Registered: February 2008

Member

Yeah, thanks Smile

My trouble was not how modify the values but find a properly policy to assign resources. For example, as I say before, I have created a VE with --config split5, I mean, to divide whole phisical resources into "5" VE machines. But now,Im increasing individually one parameter of one VE so the policy of divide resources into 5 same VE is "not correct".

Before
#PRIVVMPAGES="155119:170630"
After
PRIVVMPAGES="262144:275000"

Now I have increase the total value of RAM to 2^20. I hope this avoid the latest errors this VE. When i noticed the errors of this VE, its was unable to access into shell of Ve. I had to stop and start VE to enter by ssh.

Thanks!

Report message to a moderator

Previous Topic:	DNS wont work and cant ping but it works...
Next Topic:	Prioritizing VPS

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sun May 12 04:45:31 GMT 2024

Total time taken to generate the page: 0.01623 seconds