Re: Linux Performance Issue
On Wed, 17 Nov 2004 11:43:15 -0500, Post, Mark K <[EMAIL PROTECTED]> wrote: > One thing you're "missing" is that by using that export DISPLAY command, > you're not using the SSH tunnel, and losing the value of the encryption that > SSH provides. And you might want to look into the client configuration that tells which symmetric encryption algorithms to try first with the host. I found that the default 3des is burning much more CPU than Blowfish. For one of my presentations I compared SuSE install with and without a GUI. The conclusion was that the extra resources used to install with GUI were worth several months of running an idle server. And these resources are taken off your production capacity, probably during office hours. -- Rob van der Heij rvdheij @ gmail.com -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of John Kaba Sent: Tuesday, November 16, 2004 5:18 PM To: [EMAIL PROTECTED] Subject: Re: Linux Performance Issue Update on the SuSE SLES 8 performance Issue: KDE was just turned on to configure the various components(DB2, etc), we are using runlevel 3. I have since turned KDE off, I have reduced the Virtual memory on the Linux guest machine to 384MB, and set timer ticks to 0. I shutdown Linux, logged off the guest, and reautologged, and didn't even start the directory server, or DB2. -snip- but it is still our biggest CPU hog (52%). Of course it's not really that big of a deal with our system indicate at: ind AVGPROC-012% 01 XSTORE-00/SEC MIGRATE-/SEC MDC READS-34/SEC WRITES-01/SEC HIT RATIO-084% STORAGE-010% PAGING-/SEC STEAL-000% Q0-2(0) DORMANT-00050 Q1-0(0) E1-0(0) Q2-0(0) EXPAN-001 E2-0(0) Q3-4(0) EXPAN-001 E3-0(0) PROC -012% LIMITED-0 What does "top" show you now? Adam -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
I don't quite understand how a guest can be using 52% of the CPU, but your system is only 12% busy, unless you're running a 4-way system. As Alan Altmark points out in another note, you might want to consider upgrading to a much more current version of z/VM. It's much, much cheaper than what you're running now, and it has had lots of changes made to enhance the running of Linux guests. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of John Kaba Sent: Tuesday, November 16, 2004 5:18 PM To: [EMAIL PROTECTED] Subject: Re: Linux Performance Issue Update on the SuSE SLES 8 performance Issue: KDE was just turned on to configure the various components(DB2, etc), we are using runlevel 3. I have since turned KDE off, I have reduced the Virtual memory on the Linux guest machine to 384MB, and set timer ticks to 0. I shutdown Linux, logged off the guest, and reautologged, and didn't even start the directory server, or DB2. -snip- but it is still our biggest CPU hog (52%). Of course it's not really that big of a deal with our system indicate at: ind AVGPROC-012% 01 XSTORE-00/SEC MIGRATE-/SEC MDC READS-34/SEC WRITES-01/SEC HIT RATIO-084% STORAGE-010% PAGING-/SEC STEAL-000% Q0-2(0) DORMANT-00050 Q1-0(0) E1-0(0) Q2-0(0) EXPAN-001 E2-0(0) Q3-4(0) EXPAN-001 E3-0(0) PROC -012% LIMITED-0 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
One thing you're "missing" is that by using that export DISPLAY command, you're not using the SSH tunnel, and losing the value of the encryption that SSH provides. If "X forwarding" is turned on in the /etc/ssh/sshd_config file on the Linux system, and you use a SSH client that supports it, the DISPLAY variable should already be set when you sign on. It will look like "localhost:11" or something like that, but that is what you want. If it's not set at all, then you would need to look at the sshd_config file, or make sure you have your client set to use it. Or, use a different client. In terms of not having any X software running on your Linux system, you're not missing anything. You're absolutely doing the right thing there. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Kohrs, Steven Sent: Wednesday, November 17, 2004 9:46 AM To: [EMAIL PROTECTED] Subject: Re: Linux Performance Issue On Tue, 2004-11-16 at 16:24, Tom Duerbusch wrote: > But after the product is installed, then bring down KDE! > > I think a year or so ago, DB2/UDB required a gui screen for the > install (just to ask one question). > What am I missing here? I run X applications (DB2 installs, GSKit iKeyman, ldapxcfg) all the time from z/VM Linux guests. export DISPLAY=192.168.0.1:0.0 Now, I'm also trying to learn all the command line equivalents so I DON'T have to run X applications. :) -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
Perhaps appropriate tuning of sshd and using ssh -X would do the trick as
well, as long as you are running from a Linux w/ an X display (like from a
Knoppix boot).
The advantage there is that the sshd daemon will automagically funnel
localhost:10 (or however you tune sshd's configuration) to the client
system.
Tunnels... they're not just for crossing the Hudson River any more.
John R. Campbell, Speaker to Machines (GNUrd) {813-356|697}-5322
Adsumo ergo raptus sum
MacOS X: Because making Unix user-friendly was easier than debugging
Windows.
Red Hat Certified Engineer (#803004680310286)
IBM Certified: IBM AIX 4.3 System Administration, System Support
- Forwarded by John Campbell/Tampa/IBM on 11/17/2004 11:00 AM -
"Kohrs, Steven"
<[EMAIL PROTECTED]>To:
[EMAIL PROTECTED]
Sent by: Linux oncc:
390 Port Subject: Re: [LINUX-390] Linux
Performance Issue
<[EMAIL PROTECTED]
IST.EDU>
11/17/2004 10:45
AM
Please respond to
Linux on 390 Port
On Tue, 2004-11-16 at 16:24, Tom Duerbusch wrote:
> But after the product is installed, then bring down KDE!
>
> I think a year or so ago, DB2/UDB required a gui screen for the install
> (just to ask one question).
>
What am I missing here? I run X applications (DB2 installs, GSKit
iKeyman, ldapxcfg) all the time from z/VM Linux guests.
export DISPLAY=192.168.0.1:0.0
Now, I'm also trying to learn all the command line equivalents so I
DON'T have to run X applications. :)
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or
visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
On Wednesday, 11/17/2004 at 09:21 EST, "Mrohs, Ray" <[EMAIL PROTECTED]> wrote: > The only time we saw idle SLES8 instances hog a system is when we brought up our > z/VM 4.3 and it's guests under a 4.2 1st level at our DR site. 12 Linux > instances > pegged the CPU at 100%. We replaced the 1st level CP nucleus with 4.3, and then > everything was normal. We didn't investigate what the incompatibility was, but > it shows that some seemingly innocuous combinations of software can cause strange > behavior. Ray, I wouldn't characterize the choice of CP level as an "innocuous combination". Each release of z/VM since z/VM 4.2 has had focus on improving the performance of Linux guests. Just as with hardware, the choice of CP level should be a key element of a DR configuration. Alan Altmark z/VM Development IBM Endicott -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
On Tue, 2004-11-16 at 16:24, Tom Duerbusch wrote: > But after the product is installed, then bring down KDE! > > I think a year or so ago, DB2/UDB required a gui screen for the install > (just to ask one question). > What am I missing here? I run X applications (DB2 installs, GSKit iKeyman, ldapxcfg) all the time from z/VM Linux guests. export DISPLAY=192.168.0.1:0.0 Now, I'm also trying to learn all the command line equivalents so I DON'T have to run X applications. :) -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
The only time we saw idle SLES8 instances hog a system is when we brought up our z/VM 4.3 and it's guests under a 4.2 1st level at our DR site. 12 Linux instances pegged the CPU at 100%. We replaced the 1st level CP nucleus with 4.3, and then everything was normal. We didn't investigate what the incompatibility was, but it shows that some seemingly innocuous combinations of software can cause strange behavior. Ray Mrohs Energy Information Administration U.S. Department of Energy -Original Message- From: John Kaba [mailto:[EMAIL PROTECTED] Sent: Tuesday, November 16, 2004 3:58 PM To: [EMAIL PROTECTED] Subject: Linux Performance Issue We have just recently installed SuSE SLES8 under our VM3.1.0 system, and are experiencing some performance issues. Our Realtime monitor reveals that Linux is utilizing approx 65% of our CPU cycles, but we really have nothing running. I was told by our Linux guy to give it lots of memory, so I defined it with 1G. I'm wondering if this is my problem, or if I might have something else set up wrong. We have Tivoli Directory Server 5.2 fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not configured yet. We have just installed these products, and they are not being used at all, still we are seeing results like this: - - - Here is our Realtime Monitor results: <>z/VM CPU9672 SERIAL 0696922G DATE 11/16/04 START 13:37:27 END 14:25:02<> * USERID-> LOGGED %ACT %PGW %IOW %SUS %RUN %ELG :DSK :XST :SPL :CPU :PAG :I/O :STR *TOTALS* 5544 14.7 .0 .0 75.6 24.3.0 100100 100100100 100100 LINUX 192 71.3 .0.043.7 56.2.0 .0100 .065.8 .0 30.9 55.7 VSE2 96 100 .0 .0 90.6 9.3.0 .0.01.714.5 .0 12.1 2.9 TCPIP 96 100 .0.0 100.0.0 .0.0 .0 7.8 .0 30.7 .9 VSEIPO96 100 .0.098.9 1.0.0 .0.0 11.05.5 .0 23.7 1.1 - - - Here is a "TOP" listing from our Linux guest: 70 processes: 68 sleeping, 2 running, 0 zombie, 0 stopped CPU0 states: 1.4% user, 4.4% system, 0.0% nice, 93.0% idle CPU1 states: 14.2% user, 3.4% system, 0.0% nice, 81.2% idle Mem: 1008504K av, 953116K used, 55388K free, 0K shrd, 138316K buff Swap: 719896K av, 4K used, 719892K free 668240K cached PID USERPRI NI SIZERSS SHARE STAT %CPU %MEM TIME COMMAND 10871 dmelende150 1137211M 10632 S 2.9 1.1 7:50 kdeinit 10905 dmelende150 1680416M 14572 S 2.5 1.6 3:26 kdeinit 23343 jkaba 150 10361032 828 R 1.50.1 0:00 top 10914 dmelende150 1265212M 11724 S 0.5 1.2 0:28 kdeinit 10890 dmelende150 1298012M 11848 R 0.3 1.2 1:20 kdeinit 23327 jkaba 150 23762376 2208S 0.30.2 0:00sshd 15843 root15 0 20842080 1520S 0.1 0.2 0:06 db2fmcd 1 root15 0 208 204 160 S 0.0 0.0 0:04 init 2 root0K 0 0 00 SW 0.00.0 0:00 migration_CPU0 3 root0K 0 0 0 0 SW 0.00.0 0:00 migration_CPU1 4 root25 0 0 00 SW 0.00.0 0:00 kmcheck 5 root15 0 0 0 0 SW 0.00.0 0:00 keventd 6 root34 19 0 0 0 SWN 0.00.0 8:50 ksoftirqd_CPU0 7 root34 19 0 0 0 SWN 0.00.0 8:36 ksoftirqd_CPU1 8 root15 0 0 0 0 SW 0.00.0 0:07 kswapd 9 root25 0 0 0 0 SW 0.00.0 0:00 bdflush [EMAIL PROTECTED]:~> -
Re: Linux Performance Issue
I use a GUI installer for many MANY things and I never have used KDE or
GNOME or any other full blown desktop environment. That's like using a
howitzer to hunt pigeons. use the vncserver command.
To create and x windows session with enough color depth to display most GUI
stuff use something like:
vncserver -depth 24 -geometry 1024x768 :0
And to kill session :0 when done:
vncserver -kill :0
You can then install a vnc client (vncviewwer is what I use) to aim at your
target image and get the gui.
I've also used tight VNC, which only ships you the pixels that have
changed, rather than the whole screen. MIght be a better choice.
You can also just aim a web browser at your linux guest on port 580{port in
vncserver command} which in this case would be 0. http://{guestname}:5800
I've found the browser session method a little less stable than using
vncviewer, but your mileage may vary.
-J
Tom Duerbusch
<[EMAIL PROTECTED]
iscity.com>To
Sent by: Linux on [EMAIL PROTECTED]
390 Port cc
<[EMAIL PROTECTED]
IST.EDU> Subject
Re: Linux Performance Issue
11/16/2004 04:24
PM
Please respond to
Linux on 390 Port
<[EMAIL PROTECTED]
IST.EDU>
KDE?
Yep, you can run it, but everytime I bring up KDE, it takes about 40%
of our 60 MIP processor. There is very little that you ever need a
desktop manager for that can't be done via command line. But it is
really nice to have as a fall back to install some product.
But after the product is installed, then bring down KDE!
I think a year or so ago, DB2/UDB required a gui screen for the install
(just to ask one question).
Tom Duerbusch
THD Consulting
>>> [EMAIL PROTECTED] 11/16/04 03:14PM >>>
On Nov 16, 2004, at 2:58 PM, John Kaba wrote:
> We have just recently installed SuSE SLES8 under our VM3.1.0 system,
> and
> are experiencing some performance issues. Our Realtime monitor
reveals
> that Linux is utilizing approx 65% of our CPU cycles, but we really
> have
> nothing running. I was told by our Linux guy to give it lots of
> memory, so
> I defined it with 1G. I'm wondering if this is my problem, or if I
> might
> have something else set up wrong. We have Tivoli Directory Server
5.2
> fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not
> configured
> yet. We have just installed these products, and they are not being
> used at
> all, still we are seeing results like this:
>
You Linux image is way too big and you're probably seeing VM getting
tired paging it in and out. Try reducing it to 384MB. You will also
want to turn off timer ticks. I don't remember exactly what it is you
need to echo to where in /proc, but I'm sure someone else does.
The other thing that's hurting you is KDE. Why are you running a
desktop manager on your 390? Turn *off* X entirely--you don't have a
local display there anyway. Again, I don't remember what runlevel
"services, no X" (I think it's "2", but I won't swear to that) is in
SuSE, but I'm sure someone does.
Adam
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390
or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or
visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
On Tue, Nov 16, 2004 at 04:24:23PM -0600, Tom Duerbusch wrote: > KDE? > > Yep, you can run it, but everytime I bring up KDE, it takes about 40% > of our 60 MIP processor. There is very little that you ever need a > desktop manager for that can't be done via command line. But it is > really nice to have as a fall back to install some product. > > But after the product is installed, then bring down KDE! > > I think a year or so ago, DB2/UDB required a gui screen for the install > (just to ask one question). > > Tom Duerbusch > THD Consulting > > >>> [EMAIL PROTECTED] 11/16/04 03:14PM >>> > On Nov 16, 2004, at 2:58 PM, John Kaba wrote: > > > We have just recently installed SuSE SLES8 under our VM3.1.0 system, > > and > > are experiencing some performance issues. Our Realtime monitor > reveals > > that Linux is utilizing approx 65% of our CPU cycles, but we really > > have > > nothing running. I was told by our Linux guy to give it lots of > > memory, so > > I defined it with 1G. I'm wondering if this is my problem, or if I > > might > > have something else set up wrong. We have Tivoli Directory Server > 5.2 > > fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not > > configured > > yet. We have just installed these products, and they are not being > > used at > > all, still we are seeing results like this: > > > You Linux image is way too big and you're probably seeing VM getting > tired paging it in and out. Try reducing it to 384MB. You will also > want to turn off timer ticks. I don't remember exactly what it is you > need to echo to where in /proc, but I'm sure someone else does. echo 0 > /proc/sys/kernel/hz_timer should do it. > > The other thing that's hurting you is KDE. Why are you running a > desktop manager on your 390? Turn *off* X entirely--you don't have a > local display there anyway. Again, I don't remember what runlevel > "services, no X" (I think it's "2", but I won't swear to that) is in > SuSE, but I'm sure someone does. Have a look at /etc/inittab. There is a short doku. What you want to have is level 3 (Full multiuser with network). To change this permanently change the number in the line "id:5:initdefault:" to the runlevel you like. You can also use the yast2 runlevel-editor. To turm off the local X-Server go into /usr/X11/lib/X11/xdm/Xservers and comment out the line ":0 local /usr/X11R6/bin/X :0 vt07" Ihno -- Best regards/Mit freundlichen Grüßen Ihno Krumreich "Never trust a computer you can lift." -- Ihno Krumreich[EMAIL PROTECTED] SUSE LINUX Products GmbH Projectmanager S390 & zSeries Maxfeldstr. 5 +49-911-74053-439 D-90409 Nürnberg http://www.suse.de -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
KDE? Yep, you can run it, but everytime I bring up KDE, it takes about 40% of our 60 MIP processor. There is very little that you ever need a desktop manager for that can't be done via command line. But it is really nice to have as a fall back to install some product. But after the product is installed, then bring down KDE! I think a year or so ago, DB2/UDB required a gui screen for the install (just to ask one question). Tom Duerbusch THD Consulting >>> [EMAIL PROTECTED] 11/16/04 03:14PM >>> On Nov 16, 2004, at 2:58 PM, John Kaba wrote: > We have just recently installed SuSE SLES8 under our VM3.1.0 system, > and > are experiencing some performance issues. Our Realtime monitor reveals > that Linux is utilizing approx 65% of our CPU cycles, but we really > have > nothing running. I was told by our Linux guy to give it lots of > memory, so > I defined it with 1G. I'm wondering if this is my problem, or if I > might > have something else set up wrong. We have Tivoli Directory Server 5.2 > fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not > configured > yet. We have just installed these products, and they are not being > used at > all, still we are seeing results like this: > You Linux image is way too big and you're probably seeing VM getting tired paging it in and out. Try reducing it to 384MB. You will also want to turn off timer ticks. I don't remember exactly what it is you need to echo to where in /proc, but I'm sure someone else does. The other thing that's hurting you is KDE. Why are you running a desktop manager on your 390? Turn *off* X entirely--you don't have a local display there anyway. Again, I don't remember what runlevel "services, no X" (I think it's "2", but I won't swear to that) is in SuSE, but I'm sure someone does. Adam -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
There are some very helpful tips for running Linux guest on VM available at: http://www.vm.ibm.com/perf/tips/ Kurt Acker PatrickOBrien obrien <[EMAIL PROTECTED]> Sent by: Linux on 390 Port <[EMAIL PROTECTED]> 11/16/2004 04:46 PM Please respond to Linux on 390 Port To [EMAIL PROTECTED] cc Subject Re: Linux Performance Issue We are now messing with VM Page Space. I too have see major slowness with my SLES 8 Guest. Our VM guy has increased the Page Space, I'm checking it out now. - Original Message - From: "John Kaba" <[EMAIL PROTECTED]> To: [EMAIL PROTECTED] Subject: Linux Performance Issue Date: Tue, 16 Nov 2004 14:58:00 -0600 > > We have just recently installed SuSE SLES8 under our VM3.1.0 system, and > are experiencing some performance issues. Our Realtime monitor reveals > that Linux is utilizing approx 65% of our CPU cycles, but we really have > nothing running. I was told by our Linux guy to give it lots of memory, so > I defined it with 1G. I'm wondering if this is my problem, or if I might > have something else set up wrong. We have Tivoli Directory Server 5.2 > fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not configured > yet. We have just installed these products, and they are not being used at > all, still we are seeing results like this: > -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
Update on the SuSE SLES 8 performance Issue: KDE was just turned on to configure the various components(DB2, etc), we are using runlevel 3. I have since turned KDE off, I have reduced the Virtual memory on the Linux guest machine to 384MB, and set timer ticks to 0. I shutdown Linux, logged off the guest, and reautologged, and didn't even start the directory server, or DB2. TOP shows mostly idle: CPU0 states: 1.0% user, 0.5% system, 0.0% nice, 97.4% idle CPU1 states: 13.4% user, 0.4% system, 0.0% nice, 85.1% idle Mem: 385144K av, 79608K used, 305536K free, 0K shrd, 19536K buff Swap: 719896K av, 0K used, 719896K free 39004K cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND 784 root 15 0 1020 1016 824 R 0.9 0.2 0:00 top 1 root 15 0 236 236 196 S 0.0 0.0 0:00 init 2 root 0K 0 00 0 SW0.0 0.0 0:00 migration_CPU0 3 root 0K 0 00 0 SW0.0 0.0 0:00 migration_CPU1 4 root 25 0 00 0 SW0.0 0.0 0:00 kmcheck 5 root 15 0 00 0 SW0.0 0.0 0:00 keventd 6 root 34 19 00 0 RWN 0.0 0.0 0:00 ksoftirqd_CPU0 7 root 34 19 00 0 SWN 0.0 0.0 0:00 ksoftirqd_CPU1 8 root 25 0 00 0 SW0.0 0.0 0:00 kswapd 9 root 25 0 00 0 SW0.0 0.0 0:00 bdflush 10 root 15 0 00 0 SW0.0 0.0 0:00 kupdated 11 root 25 0 00 0 SW0.0 0.0 0:00 kinoded 12 root 25 0 00 0 SW0.0 0.0 0:00 mdrecoveryd 23 root 15 0 00 0 SW0.0 0.0 0:00 kreiserfsd 47 root 0 -20 00 0 SW< 0.0 0.0 0:00 lvm-mpd 285 root 15 0 656 656 552 S 0.0 0.1 0:00 syslogd but it is still our biggest CPU hog (52%). Of course it's not really that big of a deal with our system indicate at: ind AVGPROC-012% 01 XSTORE-00/SEC MIGRATE-/SEC MDC READS-34/SEC WRITES-01/SEC HIT RATIO-084% STORAGE-010% PAGING-/SEC STEAL-000% Q0-2(0) DORMANT-00050 Q1-0(0) E1-0(0) Q2-0(0) EXPAN-001 E2-0(0) Q3-4(0) EXPAN-001 E3-0(0) PROC -012% LIMITED-0 but we are not in production yet either. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
We are now messing with VM Page Space. I too have see major slowness with my SLES 8 Guest. Our VM guy has increased the Page Space, I'm checking it out now. - Original Message - From: "John Kaba" <[EMAIL PROTECTED]> To: [EMAIL PROTECTED] Subject: Linux Performance Issue Date: Tue, 16 Nov 2004 14:58:00 -0600 > > We have just recently installed SuSE SLES8 under our VM3.1.0 system, and > are experiencing some performance issues. Our Realtime monitor reveals > that Linux is utilizing approx 65% of our CPU cycles, but we really have > nothing running. I was told by our Linux guy to give it lots of memory, so > I defined it with 1G. I'm wondering if this is my problem, or if I might > have something else set up wrong. We have Tivoli Directory Server 5.2 > fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not configured > yet. We have just installed these products, and they are not being used at > all, still we are seeing results like this: > > --- > Here is our Realtime Monitor results: > > <>z/VM CPU9672 SERIAL 0696922G DATE 11/16/04 START 13:37:27 END > 14:25:02<> >* > USERID-> LOGGED %ACT %PGW %IOW %SUS %RUN %ELG :DSK :XST :SPL :CPU :PAG :I/O > :STR > *TOTALS* 5544 14.7 .0 .0 75.6 > 24.3.0 100100 100100100 100100 > LINUX 192 71.3 .0.043.7 > 56.2.0 .0100 .065.8 .0 30.9 55.7 > VSE2 96 100 .0 .0 > 90.6 9.3.0 .0.01.714.5 .0 > 12.1 2.9 > TCPIP 96 100 .0.0 > 100.0.0 .0.0 .0 7.8 .0 > 30.7 .9 > VSEIPO96 100 .0.098.9 > 1.0.0 .0.0 11.05.5 .0 23.7 1.1 > > --- > Here is a "TOP" listing from our Linux guest: > > 70 processes: 68 sleeping, 2 running, 0 zombie, 0 stopped > CPU0 states: 1.4% user, 4.4% system, 0.0% nice, 93.0% idle > CPU1 states: 14.2% user, 3.4% system, 0.0% nice, 81.2% idle > Mem: 1008504K av, 953116K used, 55388K free, 0K shrd, 138316K > buff > Swap: 719896K av, 4K used, 719892K free 668240K > cached > > PID USERPRI NI SIZERSS SHARE STAT %CPU %MEM TIME > COMMAND > 10871 dmelende150 1137211M 10632 S 2.9 > 1.1 7:50 kdeinit > 10905 dmelende150 1680416M 14572 S 2.5 > 1.6 3:26 kdeinit > 23343 jkaba 150 10361032 828 R > 1.50.1 0:00 top > 10914 dmelende150 1265212M 11724 S 0.5 > 1.2 0:28 kdeinit > 10890 dmelende150 1298012M 11848 R 0.3 > 1.2 1:20 kdeinit > 23327 jkaba 150 23762376 2208S > 0.30.2 0:00sshd > 15843 root15 0 20842080 1520S > 0.1 0.2 0:06 db2fmcd > 1 root15 0 208 204 160 > S 0.0 0.0 0:04 init > 2 root0K 0 0 00 > SW 0.00.0 0:00 migration_CPU0 > 3 root0K 0 0 0 0 > SW 0.00.0 0:00 migration_CPU1 > 4 root25 0 0 00 > SW 0.00.0 0:00 kmcheck > 5 root15 0 0 0 0 > SW 0.00.0 0:00 keventd > 6 root34 19 0 0 0 > SWN 0.00.0 8:50 ksoftirqd_CPU0 > 7 root34 19 0 0 0 > SWN 0.00.0 8:36 ksoftirqd_CPU1 > 8 root15 0 0 0 0 > SW 0.00.0 0:07 kswapd > 9 root25 0 0 0 0 > SW 0.00.0 0:00 bdflush > [EMAIL PROTECTED]:~> > --- > Here is the directory entry for the Linux Guest: > > USER LINUX xxx 1000M 2000M G > *-
Re: Linux Performance Issue
You'd want to change the initdefault setting from 5 to 3 (like RedHat, SuSE
uses 2 for multi-abuser w/ no network and 3 for multi-abuser with
networking).
One wonders how an initdefault of 5 got set w/o a framebuffer. Ouch.
There are also some other tunables to allow you to manage the "disk
caching" but I've long since lost the URL; it's somewhere on the
developerworks. It's a matter of echoing various values to various /proc
file thingies.
*SIGH*
(Imagines a Tyrranosaurus Rex carrying a row of Penguins down it's back,
arranged kind of like the armor plates of a Stegosaurus.)
BTW, anybody here still have a "Peace Love Linux" flag? I wanna get one
for my sailboat. It'd sure be better than the "Tampa Bay Buccaneers" flags
I usually see...
John R. Campbell, Speaker to Machines (GNUrd) {813-356|697}-5322
Adsumo ergo raptus sum
MacOS X: Because making Unix user-friendly was easier than debugging
Windows.
Red Hat Certified Engineer (#803004680310286)
IBM Certified: IBM AIX 4.3 System Administration, System Support
- Forwarded by John Campbell/Tampa/IBM on 11/16/2004 04:39 PM -
Adam Thornton
<[EMAIL PROTECTED]To:
[EMAIL PROTECTED]
mine.net>cc:
Sent by: Linux onSubject: Re: [LINUX-390] Linux
Performance Issue
390 Port
<[EMAIL PROTECTED]
IST.EDU>
11/16/2004 05:14
PM
Please respond to
Linux on 390 Port
On Nov 16, 2004, at 2:58 PM, John Kaba wrote:
> We have just recently installed SuSE SLES8 under our VM3.1.0 system,
> and
> are experiencing some performance issues. Our Realtime monitor reveals
> that Linux is utilizing approx 65% of our CPU cycles, but we really
> have
> nothing running. I was told by our Linux guy to give it lots of
> memory, so
> I defined it with 1G. I'm wondering if this is my problem, or if I
> might
> have something else set up wrong. We have Tivoli Directory Server 5.2
> fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not
> configured
> yet. We have just installed these products, and they are not being
> used at
> all, still we are seeing results like this:
>
You Linux image is way too big and you're probably seeing VM getting
tired paging it in and out. Try reducing it to 384MB. You will also
want to turn off timer ticks. I don't remember exactly what it is you
need to echo to where in /proc, but I'm sure someone else does.
The other thing that's hurting you is KDE. Why are you running a
desktop manager on your 390? Turn *off* X entirely--you don't have a
local display there anyway. Again, I don't remember what runlevel
"services, no X" (I think it's "2", but I won't swear to that) is in
SuSE, but I'm sure someone does.
Adam
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or
visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
Setting the kernel.hz_timer = 0 in the /etc/sysctl will probably help at least some but your probably going to have to do a detailed memory analysis to see how much memory is REALLY needed with DB2 in their 1GB might be the real number but my gut feel is thats WAY too large.. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Linux Performance Issue
One word. Well, two. KDE. Don't. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of John Kaba Sent: Tuesday, November 16, 2004 3:58 PM To: [EMAIL PROTECTED] Subject: Linux Performance Issue We have just recently installed SuSE SLES8 under our VM3.1.0 system, and are experiencing some performance issues. Our Realtime monitor reveals that Linux is utilizing approx 65% of our CPU cycles, but we really have nothing running. I was told by our Linux guy to give it lots of memory, so I defined it with 1G. I'm wondering if this is my problem, or if I might have something else set up wrong. We have Tivoli Directory Server 5.2 fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not configured yet. We have just installed these products, and they are not being used at all, still we are seeing results like this: --- Here is our Realtime Monitor results: <>z/VM CPU9672 SERIAL 0696922G DATE 11/16/04 START 13:37:27 END 14:25:02<> * USERID-> LOGGED %ACT %PGW %IOW %SUS %RUN %ELG :DSK :XST :SPL :CPU :PAG USERID-> :I/O :STR *TOTALS* 5544 14.7 .0 .0 75.6 24.3.0 100100 100100100 100100 LINUX 192 71.3 .0.043.7 56.2.0 .0100 .065.8 .0 30.9 55.7 VSE2 96 100 .0 .0 90.6 9.3.0 .0.01.714.5 .0 12.1 2.9 TCPIP 96 100 .0.0 100.0.0 .0.0 .0 7.8 .0 30.7 .9 VSEIPO96 100 .0.098.9 1.0.0 .0.0 11.05.5 .0 23.7 1.1 --- Here is a "TOP" listing from our Linux guest: 70 processes: 68 sleeping, 2 running, 0 zombie, 0 stopped CPU0 states: 1.4% user, 4.4% system, 0.0% nice, 93.0% idle CPU1 states: 14.2% user, 3.4% system, 0.0% nice, 81.2% idle Mem: 1008504K av, 953116K used, 55388K free, 0K shrd, 138316K buff Swap: 719896K av, 4K used, 719892K free 668240K cached PID USERPRI NI SIZERSS SHARE STAT %CPU %MEM TIME COMMAND 10871 dmelende150 1137211M 10632 S 2.9 1.1 7:50 kdeinit 10905 dmelende150 1680416M 14572 S 2.5 1.6 3:26 kdeinit 23343 jkaba 150 10361032 828 R 1.50.1 0:00 top 10914 dmelende150 1265212M 11724 S 0.5 1.2 0:28 kdeinit 10890 dmelende150 1298012M 11848 R 0.3 1.2 1:20 kdeinit 23327 jkaba 150 23762376 2208S 0.30.2 0:00sshd 15843 root15 0 20842080 1520S 0.1 0.2 0:06 db2fmcd 1 root15 0 208 204 160 S 0.0 0.0 0:04 init 2 root0K 0 0 00 SW 0.00.0 0:00 migration_CPU0 3 root0K 0 0 0 0 SW 0.00.0 0:00 migration_CPU1 4 root25 0 0 00 SW 0.00.0 0:00 kmcheck 5 root15 0 0 0 0 SW 0.00.0 0:00 keventd 6 root34 19 0 0 0 SWN 0.00.0 8:50 ksoftirqd_CPU0 7 root34 19 0 0 0 SWN 0.00.0 8:36 ksoftirqd_CPU1 8 root15 0 0 0 0 SW 0.00.0 0:07 kswapd 9 root25 0 0 0 0 SW 0.00.0 0:00 bdflush [EMAIL PROTECTED]:~> --- Here is the directory entry for the Linux Guest: USER LINUX xxx 1000M 2000M G *- ACCOUNT 442061 LINUX CPU 01 CPUID 11 CPU 02 CPUID 111222 IPL CMS PARM AUTOCR IUCV ANY IUCV ALLOW MACHINE ESA 10 OPTION MAINTCCW RMCHINFO SHARE REL 2000 XSTORE 32M CONSOLE 01
Re: Linux Performance Issue
On Nov 16, 2004, at 2:58 PM, John Kaba wrote: We have just recently installed SuSE SLES8 under our VM3.1.0 system, and are experiencing some performance issues. Our Realtime monitor reveals that Linux is utilizing approx 65% of our CPU cycles, but we really have nothing running. I was told by our Linux guy to give it lots of memory, so I defined it with 1G. I'm wondering if this is my problem, or if I might have something else set up wrong. We have Tivoli Directory Server 5.2 fixpack1, DB2 ver 8.2, and websphere express 5.1, but it is not configured yet. We have just installed these products, and they are not being used at all, still we are seeing results like this: You Linux image is way too big and you're probably seeing VM getting tired paging it in and out. Try reducing it to 384MB. You will also want to turn off timer ticks. I don't remember exactly what it is you need to echo to where in /proc, but I'm sure someone else does. The other thing that's hurting you is KDE. Why are you running a desktop manager on your 390? Turn *off* X entirely--you don't have a local display there anyway. Again, I don't remember what runlevel "services, no X" (I think it's "2", but I won't swear to that) is in SuSE, but I'm sure someone does. Adam -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
