Cluster Anomaly: WLS hogs CPU even when no clients

          I was just running a test and just need some comments on the following:
          I have the following hardware config:
          Browser -> Machine1(M1) -> Machine2(M2)
          M1=Solaris
          M2=NT Workstation
          .M1 runs WLS 5.1SP5 as webserver in proxy mode using
          "...weblogic.servlet.internal.HttpClusterServlet" with
          "...defaultServers=<M2>:...'
          .Basically just one machine in the WL cluster.
          .M2 runs WLS 5.1SP5.
          When I use the browser (on a 3rd machine) to hit a servlet on M2 via M1,
          everything is fine. I get the expected servlet response.
          However, when I hit the same servlet many times at a fast rate for about 10
          secs and then stop, I find that the CPU activity on M1 and M2 is about 90%
          and 50% respectively -- even AFTER I've stopped hitting the servlet. This
          unexplained CPU activity by WLS continues indefinitely. And I'm the only user
          on these machines and have verified that there are no other processes. I
          verified that it is WLS causing the CPU activity (using "top" on Solaris
          and "NT Task Manager" on NT).
          After I kill the WLS on M1, the CPU activity on M2 returns to zero.
          - What's going on?
          - What's WLS doing even when there are no requests hitting it?
          - Is this a bug?
          Seems to me that WLS goes into some sort of infinite loop communicating
          with machines in the cluster when it is stressed under the given config.
          The 2nd issue is: Why is the performance with a proxy so much
          worse than if I had hit the servlet directly on M2?
          I.e. Without the WLS proxy, I get, say 240 pages/sec.
          With the WLS proxy, I get only 60 pages/sec.
          Any clues?
          How else can I tune WLS to perform better aside from
          doing the stuff recommended in the tuning sections of
          the docs?
          Any insights will be most appreciated.
          Thanks.
          -Heng

          on nt you do a ctrl break.
          On solaris you do a kill -3 pid
          - Prasad
          heng chu wrote:
          > Pardon my ignorance, but how does one get a thread
          > dump of the server?
          >
          > Thanks.
          > -Heng
          >
          > Rob Woollen <[email protected]> wrote:
          > >Take a thread dump of the server when it is churning along and post it
          > >here.
          > >
          > >-- Rob
          > >
          > >Heng Chu wrote:
          > >>
          > >> I was just running a test and just need some comments on the following:
          > >>
          > >> I have the following hardware config:
          > >>
          > >> Browser -> Machine1(M1) -> Machine2(M2)
          > >>
          > >> M1=Solaris
          > >> M2=NT Workstation
          > >>
          > >> M1 runs WLS 5.1SP5 as webserver in proxy mode using
          > >> "...weblogic.servlet.internal.HttpClusterServlet" with
          > >> "...defaultServers=<M2>:...'
          > >> Basically just one machine in the WL cluster.
          > >> M2 runs WLS 5.1SP5.
          > >>
          > >> When I use the browser (on a 3rd machine) to hit a servlet on M2 via M1,
          > >> everything is fine. I get the expected servlet response.
          > >>
          > >> However, when I hit the same servlet many times at a fast rate for about 10
          > >> secs and then stop, I find that the CPU activity on M1 and M2 is about 90%
          > >> and 50% respectively -- even AFTER I've stopped hitting the servlet. This
          > >> unexplained CPU activity by WLS continues indefinitely. And I'm the only user
          > >> on these machines and have verified that there are no other processes. I
          > >> verified that it is WLS causing the CPU activity (using "top" on Solaris
          > >> and "NT Task Manager" on NT).
          > >> After I kill the WLS on M1, the CPU activity on M2 returns to zero.
          > >>
          > >> - What's going on?
          > >> - What's WLS doing even when there are no requests hitting it?
          > >> - Is this a bug?
          > >>
          > >> Seems to me that WLS goes into some sort of infinite loop communicating
          > >> with machines in the cluster when it is stressed under the given config.
          > >>
          > >> The 2nd issue is: Why is the performance with a proxy so much
          > >> worse than if I had hit the servlet directly on M2?
          > >> I.e. Without the WLS proxy, I get, say 240 pages/sec.
          > >> With the WLS proxy, I get only 60 pages/sec.
          > >>
          > >> Any clues?
          > >>
          > >> How else can I tune WLS to perform better aside from
          > >> doing the stuff recommended in the tuning sections of
          > >> the docs?
          > >>
          > >> Any insights will be most appreciated.
          > >>
          > >> Thanks.
          > >>
          > >> -Heng

Similar Messages

Firefox 3.6.6 hogs CPU 100% when playing embeded ogg video

it's really not content-related. I played the same content with an external player, and the cpu load varied between 12-40% with spikes to 50%.
HTML5 video support is great, the performance not so. Any hints?

Firefox 3.6.6 Windows XP - often goes to 80-100% CPU even with only one tab open (in this case the one I'm writing this message on). When this happens the only cure is to kill it with task manager and start again.

LR uses 10% cpu even when idle

I noticed that LR 3.3 consumes 10% cpu all the time even when it sits in the background (and I'm working within an other application and I have NOT started any batch)?!
After putting LR in the foreground (as active application) it consumes less then 5%?!
What is it doing??
Are there some settings to reduce/switch of background tasks?
(I already switched off memorycard detection for import, and XMP autosave)

hherbos wrote:
Thank's - it was the RC_ChangeManager Plugin! - I removed it and there's now no notaceable CPU consumption of LR when sitting in the background.
It's really annoying that LR misses a lot of really tiny things that Adobe developers could implement them alltogether in only a some days. But no - you have to install lots of plugins, which lack of the poor LR api.
Well, the SDK is actually pretty complete, and quite powerful. Lr is a heavily threaded application, and this is exposed to some extent by the SDK. For plugin authors, this is really nice. It offers a way to make very powerful additions to Lr that don't belong in the core product, but behave as if they are.
But it means that a poorly written plugin can soak up a fair amount of resources. Note, however, that a plugin cannot totally steal all resources. There is a governor on that, so the worst you will see is this sort of problem -- annoying, maybe, but not disasterous.
And remember that the vast majority of plugins will use the tasks and threads handed to it from the application, and will never exhibit these symptoms. It is only some sorts of plugins written in a specific manner that would ever impact overall application execution profile. So, you could install many plugins and never see a change in behaviour whether 1 or 10 were enabled. But install and run one that cycles hard on blocking activity and you will see this problem.
There is always a trade-off. Plugins make the application very flexible, but with this flexibility comes some risk. There is no perfect solution and there never will be. (Adobe could never implement all plugin behaviour everyone would want as first-class features in the core product -- this is impossible.)

Why is itunes using 94-97% of my cpu, even when idle?

i am running a 2.53ghz macbook pro with snow leopard, and whenever i have itunes running, it hovers around 94-97% cpu usage, even when idle. Whats wrong?

Try initially deleting the /Home/Library/Preferences/org.mozilla.firefox.plist file and the /Home/Library/Caches/Firefox/ folder. If you have installed any Firefox extensions/plug-ins be sure to upgrade them to the latest versions or uninstall them.

Iphoto hogs my cpu, even when running in the background

Hi, I can't leave iPhoto running in the background because it regularly uses 47% to 98% of my cpu (although there are times that it only uses 12%). Eventually the computer heats up to over 200 degrees (F) and freezes, forcing a hard re-start. I did all of the iPhoto maintenance tasks (rebuilding, etc.). My library does have 3000 photos, but I am running 16 GB of RAM. I monitor the CPU use via "Activity Monitor" and no other program comes anywhere close to using that much CPU. I'd appreciate your suggestions.

This is iPhoto for iOS. You should post your question in the iPhoto for the Mac (iLife) forum to get better answers.

SpeedGrade Pins CPU even when IDLE

MacBook Pro 2009, Core2Duo, GeForce 9600M GT, 8GB RAM
Start up SpeeGrade, all is normal.
Add a clip to the timeline, the CPU usage jumps to 200% (both cores) and stays there forever, until I delete the timeline.
Any ideas?

hherbos wrote:
Thank's - it was the RC_ChangeManager Plugin! - I removed it and there's now no notaceable CPU consumption of LR when sitting in the background.
It's really annoying that LR misses a lot of really tiny things that Adobe developers could implement them alltogether in only a some days. But no - you have to install lots of plugins, which lack of the poor LR api.
Well, the SDK is actually pretty complete, and quite powerful. Lr is a heavily threaded application, and this is exposed to some extent by the SDK. For plugin authors, this is really nice. It offers a way to make very powerful additions to Lr that don't belong in the core product, but behave as if they are.
But it means that a poorly written plugin can soak up a fair amount of resources. Note, however, that a plugin cannot totally steal all resources. There is a governor on that, so the worst you will see is this sort of problem -- annoying, maybe, but not disasterous.
And remember that the vast majority of plugins will use the tasks and threads handed to it from the application, and will never exhibit these symptoms. It is only some sorts of plugins written in a specific manner that would ever impact overall application execution profile. So, you could install many plugins and never see a change in behaviour whether 1 or 10 were enabled. But install and run one that cycles hard on blocking activity and you will see this problem.
There is always a trade-off. Plugins make the application very flexible, but with this flexibility comes some risk. There is no perfect solution and there never will be. (Adobe could never implement all plugin behaviour everyone would want as first-class features in the core product -- this is impossible.)

Photoshop CS2 hogs RAM even when all the documents are closed.

Like the title says. I was able to re-create the issue in this video: (nevermind, I guess I can't post links)
Long story short; after having documents open, I have to close Photoshop and re-open it to get my RAM back.
Why does this happen?
Thanks in advance!

The answer is given above; that is just how it works. It shouldn't really affect your system performance much if you have enough virtual memory configured, and if you've not overdone it on the amount of RAM you allow Photoshop to use.
I just close Photoshop when I'm done using it.
-Noel

30% CPU consumption even when Logic is passive

I run Logic 9.1.7 on Snow Leopard. I have an issue with my fan going full speed most of the time and when I check the Activity Monitor I can see that Logic consumes more than 30% CPU even when it's not playing (or loading or anything else). I can understand that it will consume RAM as long as it's open, but CPU?
Not that it should matter, but for the record I can add that it's a very small arrangement. Around 12 audio tracks, one exs24, some Logic effects and some Waves effects.
When I closed the song completely it went down to 18%.
Any ideas?

I run Logic under 10.7.3 on a 2011 MacBook Air 13" with i5 1.7ghz processors, 4GB RAM and 256GB SSD. I am using an external 23" Apple Cinema Display, and if I close the lid of my MBA, the fans will kick in when I run Logic. If I leave it open, the fans rarely kick into high. I have two USB mics, an Alesis iO2 USB interface and a Korg microKONTROL2 control surface configured for Logic.
My projects are pretty basic, two microphones, a few MIDI instruments... I have a one here with 8 tracks now that has a bunch of plug-ins/effects 5 of the tracks. Just did a few quick experiments using that project. When I first open it, without playing or recording anything, I see 10%-11% CPU if I click on the Activity Monitor window (ie: Logic is in the background). If I then click inside the Logic window, CPU spikes up to 33% for a moment, then returns to about 13% (ie: Logic is the active window).
Now if I hit the play button with all 8 tracks enabled, Logic uses between 43%-46% CPU. If I mute 5 of the tracks so that only 3 are enabled, it doesn't seem to matter much. However, if I play another part of the same project where only 3 of the tracks have been recorded on, then CPU drops to 35%-38%.
Now if I arm one of the microphones for recording, I see CPU use increase a bit even though I'm not actually recording. If I arm a second mic, it increases a little more, up to a total between 19%-24% when the Logic window is active. If I click in another window and put Logic in the background, CPU drops to about 18%.
If I actually record using those two mics with Logic in the active windows, CPU increases to 31%-35%.
I don't know if this helps, but you got me interested enough to try a few tests of my own. I didn't see a mention of what spec computer you are running - that could make a difference. It's also possible that MacOSX 10.7 makes a difference.

High CPU usage when in Gmail's standard HTML mode

Issue: High CPU usage (> 50%, often > 90%) ONLY when Gmail page is open. When page is closed or if I switch to basic HTML, CPU usage drops to less than 10%.
Macbook | OS X 10.6.3 | 2.4 GHz Intel Core Duo | 4 GB RAM | FF 3.63
== URL of affected sites ==
http://mail.google.com

I have a similar on my son's account on when he play games spel.nl. I think the problem is a Flash player that doesn't let go of the CPU, even when another user has the screen.

High CPU usage when zooming in TeXworks

Hello,
I have been using TeXworks for a long time in Linux (Ubuntu and Debian) and Windows and always it worked fine, but now in Archlinux when I zoom in the viewer window the magnifier is sluggish and TeXworks uses too much CPU... I was thinking that this is a problem of TeXworks, poppler, Qt or X11 but then I tried an old build and old revision of TeXworks (built in Ubuntu, to run it in Arch I have to soft link some libraries) and the zooming works flawless, so I build that source tree in Arch but the binary I get has the same problem than the other versions built in Arch, so I think it is a problem of compiler flags or the compiler itself, so...
How can I solve this?

I have a similar on my son's account on when he play games spel.nl. I think the problem is a Flash player that doesn't let go of the CPU, even when another user has the screen.

It has been only about a month that i bought my macbook pro 15 but it is heating up really fast on the left side even when the cpu usage is only 10%...i'm worried if this is a problem of the hardware.

it has been only about a month that i bought my macbook pro 15 but it is heating up really fast on the left side even when the cpu usage is only 10%...i'm worried if this is a problem of the hardware?

The way something "feels" is very subjective. We would need to know your actual temps. You can install the free iStat Pro widget which is available as a download directly from Apple:
http://www.apple.com/downloads/dashboard/status/istatpro.html
Let us know what CPU temps you're seeing at the time you feel your machine is hot. As you indicated your CPU usage was only 10%, that would imply that you already know that heavy CPU usage would be one cause for your machine to be running on the warm side... but that doesn't seem to be the cause here.

Message server hogs CPU when clients are disconnected

Hi, I have a problem where java process for the Sun One Message Queue (3.0.1) hogs the cpu when a client is disconnected. I am trying to use the MQ in an environment where clients might crash and try to reconnect in an unpredictable fashion. Most of the time, when I kill the client and restart it, I do not see the problem. Sometimes, however, doing so makes the java process of mq take almsot 100% of Cpu and it does not go down unless i kill the mq process.
I am using asynchrnous message listeners with transactions, on a persisted queue, although i saw this problem when transaction was disabled.
Does anyone have any idea what might be the source of the problem? I hope it's just a matter of configuring MQ correctly, but i could figure it out. Help would be greatly appreciated.

Are you running the enterprise edition and
using round-robin queues with multiple
receivers ?
If so, you may be running into a timing related bug
in multiple receivers that has been fixed in service
pack releases of the software
You should contact sun support to discuss testing with
the lastest service pack.
If not, can you provide more information (platform,
# and type of receivers, etc) on how you are using
the product
Thanks

MDNS responder hogging CPU when connected to network

Hi all.
It seems whenever I'm connected to my work's network (either via wired or wireless) my mDNSResponder service starts hogging CPU. It generally starts at 0 and works its way up to 70-80% total CPU usage. I can kill it but within a few seconds it starts back up again.
Any idea what the problem is or how to fix it? This has never been a problem in the past, and as far as I'm aware there have been no changes to the network.
Thanks!

My first psychic guess would be a messed up .cache, but...
mDNSResponder (also known as mdnsd on some systems) is a daemon invoked at boot time to implement Multicast DNS and DNS Service Discovery.
mDNSResponder listens UDP port 5353 for Multicast DNS Query packets.
When it receives a query for which it knows an answer, mDNSResponder issues the appropriate Multicast DNS Reply packet.
mDNSResponder also performs Multicast DNS Queries on behalf of client processes, and maintains a cache of the replies.
mDNSResponder has no user-specifiable command-line argument, and users should not run mDNSResponder manually.
To examine mDNSResponder's internal state, for debugging and disagnostic purposes, send it a SIGINFO signal, and it will then dump a snapshot summary of its internal state to /var/log/system.log, e.g.
sudo killall -INFO mDNSResponder
Also, not sure if this applies to OSX or not...
http://www.linuxmanpages.com/man8/mDNSResponder.8.php

Adobe project file is hogging shared storage bandwidth even when paused

I have a particular Premiere Pro project, that when opened, completely hogs my Terrablock storage bandwidth. I have inserted a screenshot of the storage servers disk activity even when that project file is paused. I have no other project files open or file transfers going. When I close this project, all disk activity goes to zero. This happens with no other projects and I can open it from other Mac towers with the same result.
Here is an image of the disk activity with that project open and paused.

Hi Mark,
Since my original post, I have noticed at least one other project hogging the storage bandwidth. Both projects are accessing completely different files and have a varying number of sequences.
The first project has about 45 sequences with none of them over about 90 minutes in length. The sequences are referencing .mp4s and are multicammed with two shots and occasionally contain another camera laid on top of the footage. The graphics were exported as Quicktime files and are only included in some of the sequences. This project also contains some title cards along with some png and psd files. The original stringout sequences of the footage were imported into the project using Media Browser.
The second project has 13 sequences and is referencing Quicktime Proxy footage including multicams containing two different shots. Most of the sequences contain only about 10 minutes of footage at the most. There are numerous title cards throughout this project and some of the shots are scaled in a bit. This is a duplicate of the original project (which is not hogging bandwidth) and somehow started causing bandwidth problems during the edit. No sequences were imported either.
I have also noticed that both of these projects somehow make Premiere Pro 2014 unable to close. A force quit is required for both to exit out of the program or open another project.

Cpu meter run high even when stopped

Hi im running logic 9. even when stopped the cpu meter run high, even peaking at times. I have made sure all other apps are closed while running logic. Dont know what the drain is? Please help...

Do you have a virtual instrument track selected?
Try selecting an unused audio track with no plugins and see if the useage goes down when idle.
Logic usually processes all plugins for current input on one core, which can max out your cpu.
You can also try increasing your process buffer.

Cluster Anomaly: WLS hogs CPU even when no clients

Similar Messages

Maybe you are looking for