Mac OS X Server 10.6.8 Causing CPU Load Issues for Xserves

[MacRumors]

Since the release of Mac OS X 10.6.8 last week, a number of Xserve users have reported in Apple's discussion forums (via MacNN) issues with a hardware monitoring daemon known as hwmond generating significant CPU loads of up to 95%.

Member Arminhempel reports a CPU load up to 70 percent after every reboot on several Xserves he manages; Jean-Serge Remy has an Xserve Xeon showing an 84 percent hit, while InfraredAD is experiencing a 95 percent load on a 2008 quad-core Xserve, essentially wiping out an entire core. Both the combo and delta versions of the update have generated errors and, so far, the only reliable way to fix the issue is a downgrade to Mac OS X 10.6.7.

The update also appears to be causing occasional issues with shutdowns and restarts as users attempt to address the hwmond CPU load issue.

Apple has yet to offer a specific solution for the issue, merely walking users who have contacted Apple Support through some troubleshooting steps without a specific resolution in place. But given the volume of complaints appearing in discussion forums, Apple will likely have to release a patch to address the issue.

Article Link: Mac OS X Server 10.6.8 Causing CPU Load Issues for Xserves

 

[ouimetnick]

How does Apple screw something like an Xserve up? I mean this is a BUSINESS class product, and if there is nothing wrong with it, don't fix it!

 

[sux2beyou]

I got the same issue here.
I've only updated 1 of our Xserves to 10.6.8 and hwmond was taking 94% of one core (or 20% of all the CPU's).

I was able to fix this issue by following Fred Kinder advice in the linked discussion board:

Step 1 (terminal): sudo hwmond -s250 (sets the update frequency to 250 seconds) this will give the error message identified by arminhempel.

Step 2 (activity monitor) you will now see 2 threads for hwmond, both with high CPU utilization; kill each one (using force quit)

I will probably break again at reboot, but I don't reboot servers except for required updates.

 

[Anonymous Freak]

My first-gen MacBook is having similar seeming issues. It will now randomly hang most of the OS, becoming VERY unresponsive, while the fans get louder. I can bring up the Dashboard fine, and I can hover over the Dock and have the labels appear; but I can't switch between any active applications (I tried with only Finder and TextEdit open!) I can bring up the Force Quit dialog, where all Apps show 'Not responsive', but can't actually Force Quit any of them.

My only recourse is to hard power down and restart.

 

[FBCMoyock]

CPU running above 95%

I have a similar problem since I upgraded. Did some searching this morning and found out it is related to Parallels. Obviously this isn't what everyone is experiencing, but if you have Parallels, there is a workaround here
http://kb.parallels.com/en/111541

 

[blackburn]

I've experienced weird things after the 10.6.8 update. And seems that my macbook is running hotter. (But it might be another factor)

About xserve apple doesn't really care much about it.

 

[iPhobic]

My mini server has been working hotter since the latest update. Even videos on YouTube make the fan spin faster (and noisier).

 

[roadbloc]

Don't Apple test their updates anymore? Isn't this something Microsoft should do?

 

[amadtaz]

Apple hates servers

They're trying to kill off the last of the Xserves like they did the iPhone 3G with iOS4!

 

[longofest]

How does Apple screw something like an Xserve up?

I really don't mean to make this sound like I'm hating, but I've hinted at it before in another thread, and now I'll just come out and say it - Ditch apple server products. Even before Apple pulled the plug on the Xserve hardware, the quality was just abysmal compared to competing offerings (Linux, Microsoft, and the virtualized such as ESX).

Its time to plan a migration strategy, but hey... look at it as an opportunity to virtualize!

 

[atropos]

I thought it just only me, I thought I made a wrong setting on the Xserve... now here is the clue... way to go Apple!

 

[parapup]

Parallels is a well known, widely used app - if they broke it in a maintenance update Apple's QA must be piss poor, bare minimum.

I am pissed off because Apple said to improve VPN reliability with 10.6.8 but it actually made things worse for me - where previously I had minor issues (like having to try the connection twice some times), now I have it mostly broken - 2 attempts are mandatory, the smart card suddenly stops working after sleep/wake and has to be unplugged etc.

 

[peskaa]

Noticed this the other day on our Xserve cluster. The sole machine (a 2009 2.26 Quad) I used as a guinea pig for the upgrade was pegging around 95% for the process - needless to say the others are still on 10.6.7, and the problem machine is being rolled back as we speak to an older image.

To be honest, don't expect Apple to pay much attention to Xserve updates or bugs any more. I'm still looking for replacements.

 

[yellow]

I got the same issue here.
I've only updated 1 of our Xserves to 10.6.8 and hwmond was taking 94% of one core (or 20% of all the CPU's).

I was able to fix this issue by following Fred Kinder advice in the linked discussion board:

I will probably break again at reboot, but I don't reboot servers except for required updates.

Unfortunately, this did not work for me.

hwmond kept respawning..

And even more unfortunately, I did not read this thread until or know of the issue until after updating the server to 10.6.8.

 

[Winni]

How does Apple screw something like an Xserve up? I mean this is a BUSINESS class product, and if there is nothing wrong with it, don't fix it!

It is a DISCONTINUED business class product. And one that Apple never used themselves in their own data centers - they always used Sun hardware and Sun Solaris. The fact that they never ate their own dog food tells you a lot about how committed Apple are to their own business class products.

 

[peskaa]

It is a DISCONTINUED business class product. And one that Apple never used themselves in their own data centers - they always used Sun hardware and Sun Solaris. The fact that they never ate their own dog food tells you a lot about how committed Apple are to their own business class products.

They didn't use them in their data centres, but they definitely were used elsewhere. Regent Street Apple Store (flagship) had a room full of them, I was shown round when I was purchasing 30 of them for work.

Do agree though, they've given up on the enterprise market.

 

[ratsg]

Correct, it is recently discontinued, but it is still supported.

It is a DISCONTINUED business class product. And one that Apple never used themselves in their own data centers - they always used Sun hardware and Sun Solaris. The fact that they never ate their own dog food tells you a lot about how committed Apple are to their own business class products.

 

[slaytalera]

A quick visit to the developers reference and HWMOND is the HardWare Monitoring Deamon, good luck killing it for good

 

[peskaa]

A quick visit to the developers reference and HWMOND is the HardWare Monitoring Deamon, good luck killing it for good

Yeah, it won't die permanently. Rolling back to .7 is the only solution until it gets fixed.

 

[markfrautschi]

Parallels update ParallelsDesktop-parallels-en_US-6.0.12092.670880 may help

My 2007 MacPro(1,1) 4-core runs Mac OS X 10.6.8 client, not server, and was having problems with the Parallels VM (prl_vm) task swamping two cores which was accompanied by the Parallels Desktop v6 app becoming unresponsive. (I did not notice these problems under Mac OS X 10.6.7) Coincidently Parallels Desktop for Mac update US-6.0.12092.670880 was released, which appears to have solved the problem in my case. Perhaps this is related to the XServe issue reported in this thread and an update to Parallels there will help.

Parallels is a well known, widely used app - if they broke it in a maintenance update Apple's QA must be piss poor, bare minimum.

I have a similar problem since I upgraded. Did some searching this morning and found out it is related to Parallels. Obviously this isn't what everyone is experiencing, but if you have Parallels, there is a workaround here
http://kb.parallels.com/en/111541

 

[MvdM]

How does Apple screw something like an Xserve up? I mean this is a BUSINESS class product, and if there is nothing wrong with it, don't fix it!

Probably the same way as how they screw all their other business class products (Final Cut Pro, Cinema Displays, glossy Macbook Pros).

 

[peskaa]

My 2007 MacPro(1,1) 4-core runs Mac OS X 10.6.8 client, not server, and was having problems with the Parallels VM (prl_vm) task swamping two cores which was accompanied by the Parallels Desktop v6 app becoming unresponsive. (I did not notice these problems under Mac OS X 10.6.7) Coincidently Parallels Desktop for Mac update US-6.0.12092.670880 was released, which appears to have solved the problem in my case. Perhaps this is related to the XServe issue reported in this thread and an update to Parallels there will help.

Eh? Parallels has nothing to do with hwmond. Parallels has gone nowhere near my Xserve rack, and never will.

 

[codeus]

Yeah, it won't die permanently. Rolling back to .7 is the only solution until it gets fixed.

It really isn't worth rolling back over, there is a simple workaround: -

Kill until reboot: -

sudo launchctl unload /System/Library/LaunchDaemons/com.apple.hwmond.plist

Kill permanently: -

sudo launchctl unload /System/Library/LaunchDaemons/com.apple.hwmond.plist
sudo mv /System/Library/LaunchDaemons/com.apple.hwmond.plist ~/Desktop/

The only thing I have noticed since doing this is the cpu-load leds on the front of our xserves stay fixed where they were the moment hwmond was killed, not a biggie.

HTH

 

[peskaa]

It really isn't worth rolling back over, there is a simple workaround: -

Kill until reboot: -

sudo launchctl unload /System/Library/LaunchDaemons/com.apple.hwmond.plist

Kill permanently: -

sudo launchctl unload /System/Library/LaunchDaemons/com.apple.hwmond.plist
sudo mv /System/Library/LaunchDaemons/com.apple.hwmond.plist ~/Desktop/

The only thing I have noticed since doing this is the cpu-load leds on the front of our xserves stay fixed where they were the moment hwmond was killed, not a biggie.

HTH

Is it also killing the load monitors elsewhere? We do remote monitoring as the racks are tucked out of sight in a locked room. I rolled the one test bed back to 10.6.7 this morning though, finally had time to do it. Also found one of the Xserve RAIDs had had a minor HD meltdown as well...

If it's just LEDs, then I can live with that.

 

[xiaocaomeiwatch]

ads

It really isn't worth rolling back over, there is a simple workaround: -

Kill until reboot: -

sudo launchctl unload /System/Library/LaunchDaemons/com.apple.hwmond.plist

Kill permanently: -

sudo launchctl unload /System/Library/LaunchDaemons/com.apple.hwmond.plist
sudo mv /System/Library/LaunchDaemons/com.apple.hwmond.plist ~/Desktop/

The only thing I have noticed since doing this is the cpu-load leds on the front of our xserves stay fixed where they were the moment hwmond was killed, not a biggie.

HTH

very good.