Thanks
free =
total used free shared buffers cached Mem: 507584 325348 182236 0 28296 139788 -/+ buffers/cache: 157264 350320 Swap: 1043272 0 1043272
The output of slabtop taken at the same time is a printscr (I am not running any gui on the fedora machine) at this location [url=http://www.hughes-family.com/slabtop.jpg%5DSlabtop Jpeg[/url]
Kernel = 2.6.10-1.760_FC3
I have seen all the posts that state that the memory usage is normal, and I can understand growth on demand - the release of cache pages will depend on the wash methods (fifo etc). I don't understand the constant decrement of 64k off physical and 4k or 8k added to cache that happens every 4 mins (or so) without the corresponding growth in the program sizes. If the 64k was being added to cache, then I would expect it to match. Or is this the method that fedora uses to build up the physical memory.
Why does the disk spin and the machine lock up when it hits swap? If the memory is cache, then surely it won't start swapping until the cache memory is utilised.
Any help or direction as to where to look is appreciated.
Daf
DafyddHugh wrote:
Why does the disk spin and the machine lock up when it hits swap? If the memory is cache, then surely it won't start swapping until the cache memory is utilised.
Does it just go very very slowly, or does it completely freeze? As if it's freezing, then I'd suspect that there is something odd with that swap file. You could always try booting into single user mode, use swapoff -a to stop Linux using it, and mkswap to recreate it.
Normally, Linux *will* start using swap before it has exhausted its cache, simply because there will be parts of some files that will be accessed more often than parts of some program memory.
But normally, you should hardly notice light swap usage.
If Linux doesn't completely freeze, then you might like to watch the top command and see how CPU time switches between "us" (user mode: normal programs), "sy" (kernel mode: mainly dealing with disks, filesystems and hardware), "wa" (waiting for hard disks, etc.) and "id" (idle time). That might give you pointers.
James.
On Fri, 2005-02-11 at 00:53 +0000, James Wilkinson wrote:
DafyddHugh wrote:
Why does the disk spin and the machine lock up when it hits swap? If the memory is cache, then surely it won't start swapping until the cache memory is utilised.
Does it just go very very slowly, or does it completely freeze? As if it's freezing, then I'd suspect that there is something odd with that swap file. You could always try booting into single user mode, use swapoff -a to stop Linux using it, and mkswap to recreate it.
Normally, Linux *will* start using swap before it has exhausted its cache, simply because there will be parts of some files that will be accessed more often than parts of some program memory.
But normally, you should hardly notice light swap usage.
If Linux doesn't completely freeze, then you might like to watch the top command and see how CPU time switches between "us" (user mode: normal programs), "sy" (kernel mode: mainly dealing with disks, filesystems and hardware), "wa" (waiting for hard disks, etc.) and "id" (idle time). That might give you pointers.
James.
Could this be what I have been experiencing? I had to shutdown readahead and readahead_early to keep FC3 from loading up RAM and then locking me out of the system while it decided to write dirty pages to disk. From what little I know about this, I need to tweak the configuration files in /proc/sys/vm/ to come up with a happy medium. Shutting down the 'readahead twins' seemed to have settled things down...for now. I'm hoping to find someone here that has already looked at these files and has an idea what I can be done with them. Anyone?
Paul
Hey!
I am running Fedora Core 3 and I seem to be having problem with Tomcat. I am using jakarta-tomcat-5.5.7.
When I run top I can see that the java process slowly but steadily increases the memory usage %.
When I check free I can see that -/+ buffers/cache: also decreases. After some time it takes for ever to log on to the machine since it has so little memory left.
If I shut down the tomcat process all allocated memory is released but the whole chain starts over again when I restart it.
Any ideas??
Thanks a lot! //Tom
On Thu, 2005-04-07 at 00:48 +0200, donnie wrote:
Hey!
I am running Fedora Core 3 and I seem to be having problem with Tomcat. I am using jakarta-tomcat-5.5.7.
When I run top I can see that the java process slowly but steadily increases the memory usage %.
When I check free I can see that -/+ buffers/cache: also decreases. After some time it takes for ever to log on to the machine since it has so little memory left.
If I shut down the tomcat process all allocated memory is released but the whole chain starts over again when I restart it.
Any ideas??
Thanks a lot! //Tom
The patient told his doctor "It hurts when I do this." The Doctor replied, "Then don't do that."
If the java process has a memory leak and is repeatable, file a bug on bugzilla. Search first to see if it has already been reported.
Then either don't use it, or live with the pain(inconvenience) until it is fixed.
donnie wrote:
Hey!
I am running Fedora Core 3 and I seem to be having problem with Tomcat. I am using jakarta-tomcat-5.5.7.
When I run top I can see that the java process slowly but steadily increases the memory usage %.
When I check free I can see that -/+ buffers/cache: also decreases. After some time it takes for ever to log on to the machine since it has so little memory left.
If I shut down the tomcat process all allocated memory is released but the whole chain starts over again when I restart it.
Any ideas??
Thanks a lot! //Tom
Tomcat is just a servlet container. The resources it uses itself are fairly static. If a servlet within Tomcat has a memory leak then you need to address that problem. It's not something either Fedora or Tomcat can do for you.
-
James Wilkinson -
Jeff Vian -
jery_wang -
Nigel Wade -
Paul M. Bucalo