Of
Steven Hartland
Sent: Tuesday, April 11, 2006 5:21 PM
To: hlds@list.valvesoftware.com
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
I think the problem is took them by surprise as well. From what Alfred
has said it was a fix for VAC so I suspect that affinity was set on a
thread
Oh come on, this is not a simple thing to solve. If you think it is then
go start you own company, create some great games like Valve and
then support them as well and for as long as they have. Think you'll
find that not everything is quite a easy as you think it is.
- Original Message -
--
[ Picked text/plain from multipart/alternative ]
Pretty much everyone here or who hosts servers will have a relatively
different solution to how they run/allocate servers and the problems that
are likely to occur on them. CPU usage varies per map, on a map change, when
a server restarts, other
To: hlds@list.valvesoftware.com
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
Oh come on, this is not a simple thing to solve. If you think
it is then
go start you own company, create some great games like Valve and
then support them as well and for as long as they have. Think you'll
I think the problem is took them by surprise as well. From what Alfred
has said it was a fix for VAC so I suspect that affinity was set on a
thread dedicated to VAC and the unexpected side effect was that the
entire server then has affinity set.
Not 100% but that's how I read it. Yes this is a
just wish they would ban
instantly.
This would stop 97% of all hacking.
---Original Message---
From: Wayne
Date: 04/09/06 17:18:28
To: hlds@list.valvesoftware.com
Subject: RE: [hlds] CS1.6 Servers all bound to 1 CPU?
I'd go as far as to say not questionable
Fact is if you can come up with something static, I'll give you
and example where it falls flat on its face making totally the
wrong decision.
This is NOT a problem which has a static solution, any attempt
to do so would be a huge waste of time.
Steve
- Original Message -
From:
--
[ Picked text/plain from multipart/alternative ]
Well I guess if you don't know what servers you are running, and don't know
how they run, and don't know what the normal operating parameters are for
them, then yes, you will probably be making the wrong decisions, and that
would be my own stupid
LOL you really dont get it do you :( All these things a variable,
not fixed, changeable so a decision made at point X in time will
not be correct for point X + Y in time.
- Original Message -
From: Whisper [EMAIL PROTECTED]
Well I guess if you don't know what servers you are running,
--
[ Picked text/plain from multipart/alternative ]
And you seem to think we have to cover every possibility no matter how
ridiculous it may be, just because it is not impossible.
In any case we will have to agree to disagree, since you think its all too
difficult where as I think it should not
- Original Message -
From: Whisper [EMAIL PROTECTED]
[ Picked text/plain from multipart/alternative ]
And you seem to think we have to cover every possibility no matter how
ridiculous it may be, just because it is not impossible.
No just the fact that there a huge amount of basic
--
[ Picked text/plain from multipart/alternative ]
When exactly are you running more than 4-6 HLDS/SRCDS per physical server?
It's the same thing, day in day out Steven, unless you are one of those
server providers that deliberately over-subscribe their boxes.
Exactly what is it that you are
Can you say quad? Yes our current baseline box is a Dual AMD 275
so 4 * 2.2Ghz of CPU power.
If you cant see the variables after what's been said then I'm sorry
I cant help you.
Steve
- Original Message -
From: Whisper [EMAIL PROTECTED]
When exactly are you running more than 4-6
--
[ Picked text/plain from multipart/alternative ]
Well you are not the only one with Dual Core Dual CPU Opteron Servers, and I
honestly do not see how you manage to complicate the issue so much, that
manually setting the affinitys causes such a huge issue for you if you have
a decent Game Server
LOL... Okay guys... we get it, you're both very knowledgeable.
-Original Message-
From: Whisper [mailto:[EMAIL PROTECTED]
Sent: Monday, April 10, 2006 1:26 PM
To: hlds@list.valvesoftware.com
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
--
[ Picked text/plain from multipart
I do this already in my management system. I used to use XCPU.exe, but
about a year ago I switched to using my own code (which also integrates
a sandbox ... to eliminate those plugin problems that people were
complaining about -- on hlds_linux i think)
Steven Hartland wrote:
Fact is if you can
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
Date: Mon, 10 Apr 2006 10:52:19 +0100
Fact is if you can come up with something static, I'll give you
and example where it falls flat on its face making totally the
wrong decision.
This is NOT a problem which has a static solution, any attempt
to do
Servers all bound to 1 CPU?
Date: Tue, 11 Apr 2006 02:54:04 +1000
--
[ Picked text/plain from multipart/alternative ]
When exactly are you running more than 4-6 HLDS/SRCDS per physical server?
It's the same thing, day in day out Steven, unless you are one of those
server providers
Ah so your saying we need to bind them to other CPU's, and not
to allow them to use the available CPU's.
Unfortunately I dont think that's going to be an option as we're
not just talking CS on a machine here and the fact that servers
are dynamically created and moved around machines means
its
It was a major cause of being unable to connect to VAC.
- Alfred
Steven Hartland wrote:
Ah so your saying we need to bind them to other CPU's, and not
to allow them to use the available CPU's.
Unfortunately I dont think that's going to be an option as we're
not just talking CS on a machine
Grr, that one thanks for letting us know :)
Laggy servers or unstable VAC, nasty choice! Think I'd go
for unstable VAC but totally understand your reasons to not
make it generally available. Possibility for a -beta noaffinity?
Out of interest how does this affect linux? Does it affect it at
This is a windows only problem.
- Alfred
Steven Hartland wrote:
Grr, that one thanks for letting us know :)
Laggy servers or unstable VAC, nasty choice! Think I'd go
for unstable VAC but totally understand your reasons to not
make it generally available. Possibility for a -beta noaffinity?
else.
Dustin Tuft
From: Steven Hartland [EMAIL PROTECTED]
Reply-To: hlds@list.valvesoftware.com
To: hlds@list.valvesoftware.com
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
Date: Sun, 9 Apr 2006 12:23:00 +0100
Ah so your saying we need to bind them to other CPU's, and not
to allow them
If you know how to statically determine a dynamic process then please
let us all know as Im sure that would be a great breakthrough in computer
science.
Your basic solution of round robin would be unworkable as there are
too many dynamic variants. You may be running your two servers quite
waiting for something before you move on.
Dustin Tuft
From: Steven Hartland [EMAIL PROTECTED]
Reply-To: hlds@list.valvesoftware.com
To: hlds@list.valvesoftware.com
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
Date: Sun, 9 Apr 2006 22:39:07 +0100
If you know how to statically
So you augment it to change the cpu affinity based on the cpu's load
at the instant the server is started, really what use is that? Its a single
snapshot in time of a rapidly changing environment so its almost
guaranteed to make the wrong decision.
You could build static load metrics based on
effectiveness in CS1.6 is questionable anyway so I'd rather the
problem not be mine to fix.
-- dave
Original Message
From: Dustin Tuft [EMAIL PROTECTED]
To: hlds@list.valvesoftware.com
Sent: Monday, April 10, 2006 9:07 AM
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
Well I guess
anything onto the machines.
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of David Harrison
Sent: Monday, 10 April 2006 9:52 AM
To: hlds@list.valvesoftware.com
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
I wholeheartedly agree with Steve here; we're
stop 97% of all hacking.
---Original Message---
From: Wayne
Date: 04/09/06 17:18:28
To: hlds@list.valvesoftware.com
Subject: RE: [hlds] CS1.6 Servers all bound to 1 CPU?
I'd go as far as to say not questionable, but utterly pointless having VAC
on 1.6.
Clearly, IMO, VALVe have dropped VAC
.
---Original Message---
From: Wayne
Date: 04/09/06 17:18:28
To: hlds@list.valvesoftware.com
Subject: RE: [hlds] CS1.6 Servers all bound to 1 CPU?
I'd go as far as to say not questionable, but utterly pointless having VAC
on 1.6.
Clearly, IMO, VALVe have dropped VAC updating for 1.6 and with good
and will eventually be
detected by vac.
Either way you look at it, even if vac stops one hacker its a good thing.
---Original Message---
From: Stuart Stegall
Date: 04/09/06 19:50:00
To: hlds@list.valvesoftware.com
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
This is a multi-part message
@list.valvesoftware.com
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
This is a multi-part message in MIME format.
--
[ Picked text/plain from multipart/alternative ]
Try writing your own directx hooks and then see how long before you get
hacked. Of course publically available hacks
How can i check in windows xp pro task manager what cpu is used by task?
- Original Message -
From: Gary [EMAIL PROTECTED]
To: hlds@list.valvesoftware.com; hlds@list.valvesoftware.com
Sent: Saturday, April 08, 2006 8:04 AM
Subject: RE: [hlds] CS1.6 Servers all bound to 1 CPU?
I'll
This is a massive issue for hosters who rely on multi servers taking full
advantage or multi core / multi CPU performance. It would explain the
increase in complaints about lag we have had recently.
For us its critical that this is fixed for next week or we will be unable
to run our CS comp
Use one of the many programs mentioned on this list to work around the
problem for now. The full fix requires significant changes to our timing
architecture (to stop trusting windows) and isn't a quick fix.
- Alfred
Steven Hartland wrote:
This is a massive issue for hosters who rely on multi
Servers all bound to 1 CPU?
Use one of the many programs mentioned on this list to work around the
problem for now. The full fix requires significant changes to our timing
architecture (to stop trusting windows) and isn't a quick fix.
- Alfred
Steven Hartland wrote:
This is a massive issue
PM
Subject: RE: [hlds] CS1.6 Servers all bound to 1 CPU?
Use one of the many programs mentioned on this list to work around
the problem for now. The full fix requires significant changes to
our timing architecture (to stop trusting windows) and isn't a quick
fix.
- Alfred
Steven Hartland
Surely that has the same effect as removing the binding to a single
processor which would be a quick fix?
Steve
- Original Message -
From: Alfred Reynolds [EMAIL PROTECTED]
Use one of the many programs mentioned on this list to work around the
problem for now. The full fix requires
] CS1.6 Servers all bound to 1 CPU?
Surely that has the same effect as removing the binding to a single
processor which would be a quick fix?
Steve
- Original Message -
From: Alfred Reynolds [EMAIL PROTECTED]
Use one of the many programs mentioned on this list to work around
TSC has not ever really been SMP safe, but I suppose Microsoft hacked
a way to do it :)
I wish sleep(1) would actually sleep for 1, not to 1.95
At 06:07 PM 4/8/2006, Steven Hartland wrote:
Just trying to determine if using the external utils would have the
same effect
as removing the recently
No it doesn't. Those programs let you bind hlds to another CPU, but it
stays local to that CPU. Removing the call would let the app wander over
all the CPU's again and the bug would return.
- Alfred
Steven Hartland wrote:
Surely that has the same effect as removing the binding to a single
I'm seeing this on all of my servers after I updated from version 24
to 29.. they don't seem to move off of cpu0 unless I deselect it.
At 10:03 PM 4/4/2006, David Harrison wrote:
Bit of a weird problem - we have four CS (1.6) servers running on one
machine (dual Opteron, dual core). For some
This is due to a fix stopping timing related errors (one thread is now
bound to CPU0 and windows appears to interpret that as the whole program
should be...). We are going to remove the programatic binding to CPU0
and work around the windows timing problem in another (hackier) way. I
don't have an
I'll write a command line affinity tool post it to this list. :)
At 11:55 PM 4/7/2006, Alfred Reynolds wrote:
This is due to a fix stopping timing related errors (one thread is now
bound to CPU0 and windows appears to interpret that as the whole program
should be...). We are going to remove the
Pointy hat to me.
http://www.beyondlogic.org/solutions/processutil/processutil.htm
At 11:55 PM 4/7/2006, Alfred Reynolds wrote:
This is due to a fix stopping timing related errors (one thread is now
bound to CPU0 and windows appears to interpret that as the whole program
should be...). We are
This is a multi-part message in MIME format.
--
[ Picked text/plain from multipart/alternative ]
BindCPU works quite well.
http://www.softtreetech.com/24x7/extras/BindCPU.exe
All you do is start it like this bindcpu.exe 2 cod2.exe blahb
lahbkdkdfjdjk more command gobble dee gook.
For MOST
and here is setaffin (same basic thing) with sourcecode in C:
http://www.hayseed.net/~emerson/setaffin.html
- Original Message -
From: Kevin Ottalini
To: hlds@list.valvesoftware.com
Sent: Friday, April 07, 2006 9:11 PM
Subject: Re: [hlds] CS1.6 Servers all bound to 1 CPU?
bindcpu
--
[ Picked text/plain from multipart/alternative ]
I saw similar behaviour when I tried to start some servers with affinity
switch with few other server without it. As a result all servers clung to
single CPU no matter what run-time affinity setting I made manually. Problem
disappeared after I
i have several dual opterons and 1 dual core single
i dont see this behavior
have you (just as a test) set them each to one core?
dex
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] Behalf Of David Harrison
Sent: Tuesday, April 04, 2006 10:03 PM
To:
: Affordablegameservers.com [EMAIL PROTECTED]
To: hlds@list.valvesoftware.com
Sent: Wednesday, April 05, 2006 12:07 PM
Subject: RE: [hlds] CS1.6 Servers all bound to 1 CPU?
i have several dual opterons and 1 dual core single
i dont see this behavior
have you (just as a test) set them each to one core
50 matches
Mail list logo
