Click to See Complete Forum and Search --> : Strange disk activity
Hello!
I have 196 MB of RAM. Not so much to brag about, but enough to cope without a swap partition, so I do not have one. But at memory leaks (Netscape 6.1 PR1) disk activity is so high it is impossible to even move the mouse pointer. Why is that? And why isn't the kernel killing that process? Is there a way to set a "memory qouta" so the system can't run out of memory?
Thanks! :)
demian
06-20-2001, 08:27 PM
Originally posted by AE:
<STRONG>I have 196 MB of RAM. Not so much to brag about, but enough to cope without a swap partition, so I do not have one.</STRONG>
Big mistake. 512MB, heck even 1GB is not enough to do without some swap space. It's in the way the linux kernel handles memory it will not swap anything until the physical RAM is full. If that happens, however, it relies on being able to swap at least something to some designated HD space. Otherwise bad things will happen as you noticed. Do yourself a favour and create a swap file of at least 100MB. (No need to create a partition a swap file works just as well.)
edit:
You can limit the memory usage of netscape. It's somewhere in the preferences menu but that won't solve your problem.
[ 20 June 2001: Message edited by: demian ]
AE,
If you haven't already, check to make sure that you don't have any non-essential processes running in the background. Cron jobs and such can bog your system down if they kick in at the wrong time.
Big mistake guess so... I tought that I would avoid any memory related disk activity by not having a swap-partition. I didn't realize that Linux apparently have to have one. Thanks for enlighten me! :)
And thanks DMR, that could have been the problem. But when I checked when the last cron-processes was run it was about one hour before the problem occurred.
Hmmm, just a thougth (again). Would it be possible to create a small RAMdisk for use as a swap partition? :rolleyes:
demian
06-21-2001, 07:46 AM
Originally posted by AE:
<STRONG>[QUOTE]Would it be possible to create a small RAMdisk for use as a swap partition?</STRONG>
It should be possible but then that somehow beats the purpose of swap space, doesn't it? If HD space is so tight you might want to try and go with a really small swap file. I have 256MB RAM and as much swap space. The swap, however, is hardly touched (right now 1.3MB are used). So maybe you start with a 10MB swap file and push your system a bit while monitoring the memory and swap usage. That way you should be able to adjust the size of the swap file quite accurately.
It's not the disk space that I am concerned about, it's the speed. I have hard rebooted the system many times, simply because I can't stand waiting any longer after a half hour of intense disk activity (not being able to do a thing). When using only RAM it would, hopefully, take less time for the memory to run out and the kernel to send a nice "SuSE kernel: Out of Memory: Killed process 1433"-message. Am I wrong again?
demian
06-21-2001, 08:28 AM
What kind of system do you have? Esp. how many harddisk and how are they partitioned? What kind of jobs do you have running that make the disk go wild?
I really don't think a swap file on the hard drive will slow things down. I mean say there are like 10MB swapped. Reading/writing 10MB takes no time even with a crappy old ide disk.
I have a PII266, a three year old 6.4 GB IDE HD with three partitions [5,1,0.4], all ext2, and running SuSE 7.1 (on the 5 GB partition). The other two partitions is currently empty.
The software that makes my system sad is buggy with memory leaks (such as Netscape 6.1 Pre Release 1 and XMMS 1.2.3).
With quality software my system works fine. But what kind of OS would require that program never leaks memory?
I normal cases the hard disk wouldn't be too slow. But obviously, when the memory is about to be full, harddisk read and/or write is the only thing the system is capable to do. I don't think that it could be a bug in, for example, Netscapes disk cache, since it would not have that high priority.
I was just thinking that if this disk activity took place in memory instead, it would go a whole lot faster.