SLES9: Large Mmeory Consumption/Leakage?
I have a vanilla installed SLES9 system with no applications runnig as of yet. Everything is default (except I disabled the ldap server and set the default run level to 3 so no X processes are running either) However, it seems to be using a rather large amount of memory compared to other distro's. Here is a snapshot from top: (I know top is not an accurate means to getting this info but it is always relatively close) top - 23:52:38 up 1:59, 1 user, load average: 0.00, 0.00, 0.00 Tasks: 52 total, 2 running, 50 sleeping, 0 stopped, 0 zombie Cpu(s): 0.2% us, 0.0% sy, 0.0% ni, 99.8% id, 0.0% wa, 0.0% hi, 0.0% si Mem: 1034168k total, 728752k used, 305416k free, 74364k buffers Swap: 1049528k total, 0k used, 1049528k free, 516876k cached When the system first booted memory usage for around 500mb (still high) but during it's 2 hour of uptime it has grown to 728mb, by itself, with no system activity. When I look at 'ps -aux' there doesn't seem to be any single process using a large amount of memory. This system is a dual Xeon, 1gb memory, SysKonnect gigabit NIC, 2x36gb SCSI disks with 3 Raid 1 devices. (/boot, swap, and /) This just seemed like a large amount of memory being used with no applications running other than the default services. Thanks for any info, CC
On Thu, December 9, 2004 11:28 am, Rhugga said:
I have a vanilla installed SLES9 system with no applications runnig as of yet. Everything is default (except I disabled the ldap server and set the default run level to 3 so no X processes are running either)
However, it seems to be using a rather large amount of memory compared to other distro's. Here is a snapshot from top: (I know top is not an accurate means to getting this info but it is always relatively close)
Naw, it's about the same on all distros. Your system is using 728752k - 516876k - 74364k = 137512k or about 134Megs. (cache) (buffers)
top - 23:52:38 up 1:59, 1 user, load average: 0.00, 0.00, 0.00 Tasks: 52 total, 2 running, 50 sleeping, 0 stopped, 0 zombie Cpu(s): 0.2% us, 0.0% sy, 0.0% ni, 99.8% id, 0.0% wa, 0.0% hi, 0.0% si Mem: 1034168k total, 728752k used, 305416k free, 74364k buffers Swap: 1049528k total, 0k used, 1049528k free, 516876k cached
When the system first booted memory usage for around 500mb (still high) but during it's 2 hour of uptime it has grown to 728mb, by itself, with no system activity.
When I look at 'ps -aux' there doesn't seem to be any single process using a large amount of memory.
This system is a dual Xeon, 1gb memory, SysKonnect gigabit NIC, 2x36gb SCSI disks with 3 Raid 1 devices. (/boot, swap, and /)
This just seemed like a large amount of memory being used with no applications running other than the default services.
It's just cache and buffer usage which will be freed if a more important use comes up like if you load a program. Unused RAM is wasted RAM.. It's a valuable resource and Linux maximizes its use. Cheers, Sean
CC / Rhugga, On Thursday 09 December 2004 08:28, Rhugga wrote:
I have a vanilla installed SLES9 system with no applications runnig as of yet. Everything is default (except I disabled the ldap server and set the default run level to 3 so no X processes are running either)
...
When the system first booted memory usage for around 500mb (still high) but during it's 2 hour of uptime it has grown to 728mb, by itself, with no system activity.
This is as it should be. The system does not try to maintain a large margin of unused RAM, since that is nothing more than waste as a policy. Instead, it keeps a lot of discardable resources (disk blocks, mostly) in RAM and manages any RAM not required by applications as a variable-sized cache. This is a change from older Linux kernel behavior, but I'm not sure how far back the change of policy goes.
When I look at 'ps -aux' there doesn't seem to be any single process using a large amount of memory.
They're not, and that's the sign that all is well. Should one or more applications request memory, the kernel will abandon enough of the cached disk contents to fulfill those applications' requests.
This system is a dual Xeon, 1gb memory, SysKonnect gigabit NIC, 2x36gb SCSI disks with 3 Raid 1 devices. (/boot, swap, and /)
This just seemed like a large amount of memory being used with no applications running other than the default services.
The only fix required is that you adjust your intuition about a healthy system's resource utilization patterns. You want to start to worry when the swap activity goes up, not until. I have a single-CPU (w/ hyperthreading) and one gigabyte of RAM and I see very similar patterns and my machine runs very smoothly.
Thanks for any info, CC
Randall Schulz
On Thursday 09 December 2004 10:28 am, Rhugga wrote:
top - 23:52:38 up 1:59, 1 user, load average: 0.00, 0.00, 0.00 Tasks: 52 total, 2 running, 50 sleeping, 0 stopped, 0 zombie Cpu(s): 0.2% us, 0.0% sy, 0.0% ni, 99.8% id, 0.0% wa, 0.0% hi, 0.0% si Mem: 1034168k total, 728752k used, 305416k free, 74364k buffers Swap: 1049528k total, 0k used, 1049528k free, 516876k cached
CC
Try htop from http://htop.sourceforge.net instead of top... Very nice and easy way to see memory usage. Both Sean and Randall have pointed out that all is looking good on your system. htop just does top one better IMO. Stan
Stan, On Thursday 09 December 2004 10:02, Stan Glasoe wrote:
...
Try htop from http://htop.sourceforge.net instead of top... Very nice and easy way to see memory usage. Both Sean and Randall have pointed out that all is looking good on your system. htop just does top one better IMO.
Thanks for the tip. I found SuSE RPMs at http://rpm.pbone.net/. Simply search for "htop" and the SuSE RPMs are the first four in the returned search results. RPMs are available for SuSE 8.2, 9.0, 9.1 and 9.2. My first impression is that it's indeed superior to stock "top." Thanks again.
Stan
Randall Schulz
Stan, On Thursday 09 December 2004 10:20, Randall R Schulz wrote:
Stan,
On Thursday 09 December 2004 10:02, Stan Glasoe wrote:
...
Try htop from http://htop.sourceforge.net instead of top... Very nice and easy way to see memory usage. Both Sean and Randall have pointed out that all is looking good on your system. htop just does top one better IMO.
Thanks for the tip. I found SuSE RPMs at http://rpm.pbone.net/. Simply search for "htop" and the SuSE RPMs are the first four in the returned search results. RPMs are available for SuSE 8.2, 9.0, 9.1 and 9.2.
My first impression is that it's indeed superior to stock "top."
Can you suggest the best way to view htop output? For routine CLI purposes, I use Konsole with black text on a white background. This does not seem to be optimal w.r.t. the color scheme used by htop, and it does not appear that its color scheme is configurable--at least there's no mention of it on the Web site or in the man page. Thanks. Randall schulz
On Thursday 09 December 2004 12:31 pm, Randall R Schulz wrote:
Stan,
Can you suggest the best way to view htop output? For routine CLI purposes, I use Konsole with black text on a white background. This does not seem to be optimal w.r.t. the color scheme used by htop, and it does not appear that its color scheme is configurable--at least there's no mention of it on the Web site or in the man page.
Randall schulz
I use Konsole with: Font: Linux Keyboard: XTerm (XFree 4.x.x.) Schema: Green on Black (per the old Compaq luggable 286 days) My keyboard setting was critical. Using 'linux console' didn't work where 'XTerm (XFree 4.x.x)' did. 'Linux Colors', 'Transparent, Dark Background' adn 'White on Black' for schema are also OK to my eyes. Stan
On Thursday 09 December 2004 12:20 pm, Randall R Schulz wrote:
Thanks for the tip. I found SuSE RPMs at http://rpm.pbone.net/. Simply search for "htop" and the SuSE RPMs are the first four in the returned search results. RPMs are available for SuSE 8.2, 9.0, 9.1 and 9.2.
My first impression is that it's indeed superior to stock "top."
Thanks again.
Randall Schulz
Di nada, dude. That's the same rpm I got from apt/synaptic using ftp.gwdg.de as the repository. Hadn't even tried it until this thread reminded me I had loaded it a couple days ago because it sounded better than top. top is great and the htop enhancements just puts it into the "I'll always be using this one" category. Stan
On Thursday 09 December 2004 01:02 pm, Stan Glasoe wrote:
On Thursday 09 December 2004 10:28 am, Rhugga wrote:
top - 23:52:38 up 1:59, 1 user, load average: 0.00, 0.00, 0.00 Tasks: 52 total, 2 running, 50 sleeping, 0 stopped, 0 zombie Cpu(s): 0.2% us, 0.0% sy, 0.0% ni, 99.8% id, 0.0% wa, 0.0% hi, 0.0% si Mem: 1034168k total, 728752k used, 305416k free, 74364k buffers Swap: 1049528k total, 0k used, 1049528k free, 516876k cached
CC
Try htop from http://htop.sourceforge.net instead of top... Very nice and easy way to see memory usage. Both Sean and Randall have pointed out that all is looking good on your system. htop just does top one better IMO.
Stan ==========
And for those using KDE gui, a simple press of the ctrl-esc keys will open up ksysguard to give a fully graphical view of the system processes, memory usage, cpu, etc. Plus you can kill any process you want to, except for root processes of course. Lee
participants (5)
-
BandiPat -
Randall R Schulz -
Rhugga -
Sean -
Stan Glasoe