Fedora runs out of memory and locks up
 

Greetings!

I'm having trouble running down a cause (and solution) for this situation. Here's what I've got:

Fedora Core-1
1.6ghz Sempron
750mb RAM

Running Gnome

The machine boots fine with no errors. After which, I check memory by using "free". It shows about 600mb free. But the problem is, throughout about 2 days of uptime (no activity on the machine), all memory (except maybe 10mb) is consumed! The memory usage decrements a couple of K every second or so... Then, when I try to run an application ie.Firefox, the system locks up. I can't even connect to it via the network (though, it can be pinged).

My only option is to hit the power switch and start over. Otherwise, the machine runs great! This problem happens if the machine is in either 3 or 5 init state.

I've checked the process output and nothing seems out of the ordinary. Nothing seems unusual in /var/log/messages.

I'm thinking of back-reving to RH 9.0. I didn't have this problem in 9.0

Any thoughts!?

Thanks,
Bob

These two issues are likely to be unrelated. Check if you are measuring the actual unused memory, or the free memory plus cache buffers. The free memory will normally go down as more things are read/write cached from the disk; there's pleanty of posts about that on these forums.

If this really is an out of memory problem, then you should look at investing in some swap space. Most distros will set up a swap partition for you by default, but you can easily set up a swap file (or even a partition) if it didn't.

You might also want to run top and sort processes by memory usage; this will tell you which process (if any) has the memory leak. Even so, a process with leakey memory will just get itself killed eventually, it won't cause the system to lock up.

Check your /var/log/messages file after a crash to try and identify the cause. You might also need the sync option on the /var partition (or root partition, if you don't have a /var) to force the kernel to flush the log files to disk before the reboot.

Thank you for the reply!

I had not considered using "top" to view memory because of its own resource usage! However, I will certainly try it. I've been using the ps -eo options to view memory. I didn't want top to skew my research.

I do have swap space defined (in retrospect, I should have included that in the first post), for about 1.0gb. Funny thing, no swap space is utillized when the machine is near the end of its memory usage.

And for the "free" memory output, the row I'm reading from the "free" command is "Mem:" and the column heading is "free".

I have checked the /var/log/messages file and there are no entries in it for anything pertaining to the system locking up! It appears the system locks up, therefore, it can't write to the messages file.

The system will never cache hard disk in the swap (for obvious reasons) so I suspect that the apparant memory usage increase is actually a red herring (you're looking at memory used for caching; this will be freed as needed).

This leaves us with a system that locks up if you try and run firefox after it's been idle for two days; there are a number of things that could cause this ranging from overheating to DMA issues. I'd be quite surprised if turning on sync for /var wouldn't cause some debug messages to appear somewhere.

You might also try running xconsole before you leave the machine idle; errors generally get logged to the console before being logged to the disk.

Excellent ideas.

I feel confident that overheating is not an issue... perhaps there is an issue with DMA.

I'll give the "xconsole" option a try!

Thank you again!