PDA

View Full Version : 10.5: kernel_task jumping to ≥100% CPU on boot




Lawen
Jun 24, 2010, 11:03 AM
I'm trying to help a buddy out and am completely stumped. Anyone have something I can try next?

Background:
Dual 1.8 PPC Mac Pro running Server 10.5. All patches and updates installed. One large (1TB) external firewire drive attached. Wired to an Airport Base Station with a SSH port forward rule in place pushing TCP port 22 traffic to the server.

He basically uses it solely as a file server. He copies a file from the external drive to his workstation, works on it, saves it, and copies it back to the external drive on the server. I have a cron'd rsync that runs nightly and syncs his external drive with his business partner's remote server. That's all it's used for: storing files and rsync'ing the stored files.

Problem:
Since yesterday, his system has slowed to a crawl. Doing anything on the server, especially file copies to/from the workstation, gives him a spinning beachball for 5 minutes before anything happens. Activity Monitor shows the CPU running full-throttle all the time.

I've SSH'ed in and run a top, the kernel_task process is constantly using 100.2% - 100.9% of the CPU. No other processes are using more than 3 or 4% CPU. Plenty of free memory, both physical and virtual.

I've had him unmount and disconnect his external drive, then reboot. Immediately kernel_task jumps back to 100%+ CPU Util. I've had him disable the SSH port forward rule, reboot the Airport, and reboot the server; kernel_task jumps right back to 100%+. As far as I can tell, it's not an IO problem with the external drive nor a DoS attempt. I've also had him repair permissions, which doesn't seem to have helped.

The only thing of note I see in /var/log/system.log are a bunch of afctl and emond messages but as far as I can tell they're just normal portscans and shenanigans from China/Russia:
Jun 24 00:31:17 server emond[64]: Host at 201.230.18.187 will be blocked for at least 15.00 minutes
Jun 24 00:31:17 server emond[5826]: DoRunAction (child): setting the uid/gid to 0/0
Jun 24 00:31:18 server afctl[5826]: Firewall not running or managed by another entity, rule not added

I'm pretty much stumped. Anyone have any ideas or suggestions?



GDIBeavis
Jun 24, 2010, 04:58 PM
I'm trying to help a buddy out and am completely stumped. Anyone have something I can try next?

Background:
Dual 1.8 PPC Mac Pro running Server 10.5. All patches and updates installed. One large (1TB) external firewire drive attached. Wired to an Airport Base Station with a SSH port forward rule in place pushing TCP port 22 traffic to the server.

He basically uses it solely as a file server. He copies a file from the external drive to his workstation, works on it, saves it, and copies it back to the external drive on the server. I have a cron'd rsync that runs nightly and syncs his external drive with his business partner's remote server. That's all it's used for: storing files and rsync'ing the stored files.

Problem:
Since yesterday, his system has slowed to a crawl. Doing anything on the server, especially file copies to/from the workstation, gives him a spinning beachball for 5 minutes before anything happens. Activity Monitor shows the CPU running full-throttle all the time.

I've SSH'ed in and run a top, the kernel_task process is constantly using 100.2% - 100.9% of the CPU. No other processes are using more than 3 or 4% CPU. Plenty of free memory, both physical and virtual.

I've had him unmount and disconnect his external drive, then reboot. Immediately kernel_task jumps back to 100%+ CPU Util. I've had him disable the SSH port forward rule, reboot the Airport, and reboot the server; kernel_task jumps right back to 100%+. As far as I can tell, it's not an IO problem with the external drive nor a DoS attempt. I've also had him repair permissions, which doesn't seem to have helped.

The only thing of note I see in /var/log/system.log are a bunch of afctl and emond messages but as far as I can tell they're just normal portscans and shenanigans from China/Russia:
Jun 24 00:31:17 server emond[64]: Host at 201.230.18.187 will be blocked for at least 15.00 minutes
Jun 24 00:31:17 server emond[5826]: DoRunAction (child): setting the uid/gid to 0/0
Jun 24 00:31:18 server afctl[5826]: Firewall not running or managed by another entity, rule not added

I'm pretty much stumped. Anyone have any ideas or suggestions?

Try going into safe boot mode by holding down the shift key when booting. Then see if you still have the problem.

You can also do a nvram reset by going into open firmware by holding down the command option o f And then typing these commands:

reset-nvram
set-defaults
reset-all