Memory limit Redhat EL AS 3 1

[7280]

Hi,
on two redhat el as 3 systems I have installed more than 3 gb ram but both systems see 3gb. At the begnining I thought it was a hw problem, but when I saw that the limit was on both servers I immediately excluded that hypothesis.
I think I need to configure something at os level?
I hope I don't need to recompile the kernel.

Thanks in advance

 

[Ogzilal]

Hi,

How did you get that the system sees 3GB ?
What is in your /proc/meminfo file ?
Please post the result of cat /proc/meminfo

Ali

 

[7280]

grep -i memtotal /proc/meminfo
MemTotal: 3889416 kB

on both servers. I'm sure about this.
It seems I need to compile kernel for hugemem.

thanks.

 

[Ogzilal]

Hi,

Here is the same command on one of my boxes running Redhat 4 dowloaded from redhat and installed as is on a HP ML350 G4P server ( The whole physical memory is recognized at boot ) :

[root]# uname -a
Linux horus 2.6.9-34.ELsmp #1 SMP Fri Feb 24 16:54:53 EST 2006 i686 i686 i386 GNU/Linux
[root]# grep -i memtotal /proc/meminfo
MemTotal:      5191860 kB
[root]#

Ali

 

[7280]

redhat 4 uses kernel 2.6 in which huge memory is enabled by default.. in redhat 3, kernel 2.4, I don't see the correct memory. I don't know if there's some parameter to modify or to recompile kernel..

 

[Annihilannic]

RHEL AS3 works fine with systems up to 32GB, even with the stock kernel, however for more efficient memory management you should use the hugemem kernel for any system with 6GB or more RAM. You don't need to recompile the kernel, just install the kernel-hugemem RPMs (usually available from RHN using up2date) and then select the hugemem kernel in your grub.conf.

Annihilannic.

 

[7280]

So why I don't see the correct memory?
At bios level all the ram is seen correctly.

About installing hugemem it will create a new kernel, unfortunately I need then to modify it in order to suite our needs (rdac configuration).

Thanks again.

 

[7280]

Hi,
in my config-kernelversion file under /boot I have this entry:
CONFIG_3GB=y

Is it possibile that this parameter is limiting memory?

Thanks

 

[Annihilannic]

That sounds very likely to me. I've not seen that file before on AS3, but we always use the standard kernels, so I would say that's a side effect of rolling your own kernels. You must have answered that question incorrectly (or unwittingly accepted the default) when setting your build options for that kernel.

By what means do you usually configure your kernel before building it?

Annihilannic.

 

[rcarpent]

Are you using the SMP kernel. If not switch to the SMP kernel.
# grep -i memtotal /proc/meminfo
MemTotal: 5907480 kB

Linux contact 2.4.21-32.0.1.ELsmp #1 SMP Tue May 17 17:52:23 EDT 2005 i686 i686x

I am able to use the memory.
Boot off the SMC kernel and check the memory again.

 

[7280]

Hi,
yes I use the SMP kernel.
First server is: 2.4.21-20.ELsmp
Second server is: 2.4.21-27.ELsmp

I havent' installed/configured this kernel. I'm new at this job and I found systems in this state.
I was trying to understand why memory is seen correctly by bios but not by redhat.

Thanks again.

 

[7280]

I was thinking to make a copy of file /boot/config-kernelversion, comment CONFIG_3GB=Y and restart the system.
I hope kernel will boot again!!
What do you think about this?

Thanks.

 

[Annihilannic]

I doubt that will make a difference without actually recompiling the kernel.

Do you not have any contact with the person(s) who originally customised the kernel on these systems to find out why and how they did it?

Annihilannic.

 

[7280]

Unfortunately no.
I opened a case in redhat and they told me to install hugekernel even if they state that it should be done on systems with more than 16gb memory.. crazy no?
They told that CONFIG_3GB=Y doesn't concern, it is the process address space of the kernel.

 

[7280]

I forget to say that I have three linux servers, all redhat el as 3, and on all them I see only 3gb ram, even on the systems where kernel hasn't been customized.
It seems a problem with the kernel version: 2.4.21
I don't have any other explanation.

 

[Annihilannic]

2.4.21 is the only kernel version available on RHEL AS3, we use it on all of our (Dell) servers and nearly all of them have more than 4GB of RAM.

I opened a case in redhat and they told me to install hugekernel even if they state that it should be done on systems with more than 16gb memory.. crazy no?

No, not crazy. hugemem is required for systems with 16GB or more memory, but is recommended for systems with 6GB or more.

Having said that I have no idea why you can't see the rest of your memory using the smp kernel. Have you tried using hugemem at all yet?

What hardware are you using?

What is the output of cat /proc/cmdline? Are there any options on the kernel command line that might be limiting the amount of memory usable by the OS, e.g. mem=3G?

Annihilannic.

 

[7280]

cat /proc/cmdline doesn't have limits. output of 8gb server is: ro root=LABEL=/ , while output of 4gb server is: ro root=LABEL=/ ramdisk_size=15000 (ramdisk_size is required to be added when installing RDAC).
HW is:
HP ProLiant DL585 G1
IBM eserver xSeries 255 -[86855RX]

I downloaded lates hugemem kernel (kernel-hugemem-2.4.21-47.0.1.EL.i686.rpm) but it's a problem to install it for me because of the RDAC driver. Once installed I need to reinstall the RDAC driver in order to access the SAN (IBM FastT) but it's not so simple. I don't have an identical test server.

I tried to install this hugemem on my test server (a virtual machine) but it failed with this error:
No module BusLogic found for kernel 2.4.21-47.0.1.ELhugemem, aborting.
mkinitrd failed

This server has more than 4gb ram, so I wanted to see if the problem is fixed or not.

You say you have 2.4.21 and you see correct memory without installing hugemem?

Thanks again for your help.

 

[Annihilannic]

You say you have 2.4.21 and you see correct memory without installing hugemem?

Yes, here is one such example:

[tt]# uname -r
2.4.21-27.0.1.ELsmp
# free
total used free shared buffers cached
Mem: 7973704 4373108 3600596 0 266964 2409408
-/+ buffers/cache: 1696736 6276968
Swap: 2096440 0 2096440
#[/tt]

However most of our systems with this much or more memory have been upgraded to hugemem, here is one with 32GB:

[tt]# uname -r
2.4.21-47.ELhugemem
# free
total used free shared buffers cached
Mem: 32695584 32652412 43172 0 498872 17644240
-/+ buffers/cache: 14509300 18186284
Swap: 2096440 1988 2094452
#[/tt]

Annihilannic.

 

[7280]

Hi,
very strange. I don't know what kind of installation has been done here!

Redhat told me that the only solution is to install the hugemem kernel.
Can you advise how to do this?
I know that rpm -ivh is enough, but then what should I do for the SAN driver? I don't think I need to install it again since it is already installed. Maybe I need to reconfigure the kernel?
To do this as soft as possible, do you suggest the same level kernel? (eg 2.4.21-20)

Thanks again.

 

[Annihilannic]

Well, to prove to yourself whether or not it will solve your problem, why not install the kernel without the SAN module first, i.e. comment out the lines in modules.conf and install the kernel, or re-run the mkinitrd manually if it is already installed.

Reboot and check whether you can see all of the memory with the new kernel. If you can't, go back to the previous configuration, and the drawing board. If you can, try and figure out how to get your SAN modules configured for the new kernel.

Annihilannic.