| gabriel rosenkoetter on 18 Nov 2003 13:22:03 -0500 |
|
Could I ask that you guys please NOT top-reply?
It makes it more difficult to follow the thread of the conversation
not easier, because one has to jump to the bottom of a message and
read backwards to find context. Rather, it's much more standard
Netiquette to quote only the portions of the original message to
which you're directly responding *before* your response, so the
reader can have exactly the context he needs present when reading.
Never mind the massive bandwidth waste caused by top-replies.
On Tue, Nov 18, 2003 at 10:18:37AM -0500, Mattison, Jacob wrote:
> I'm running RedHat 9 on a Dell PowerEdge server that supposedly has 2 Gig of
> RAM.
Why are you running RH 9 on server hardware? (It's rather
ill-advised...)
> However, when I run "top", the total memory listed is about 1 Gig. Is
> there something I can do, perhaps kernel parameters in grub.conf or
> elsewhere, to get it to access all of the memory?
Using the bigmem, rather than the smp, kernel may be enough. You'd
have to be using a pretty old kernel version to even have a bigmem
kernel installed, though.
> Please oh please don't tell me I need to recompile the kernel..... :)
If what Chris is saying is true, then you just have to run
up2date-nox --configure, remove kernel* from the pkgSkipList, run
up2date-nox -u, then put kernel* back in the pkgSkipList.
I think that's far less painful than building your own kernel (from
RH sources or otherwise) has the potential to be, especially
considering the fact that RH (still!) ships a broken gcc(1) the last
I checked (use kgcc(1) if it exists).
On Tue, Nov 18, 2003 at 10:58:51AM -0500, Paul wrote:
> There is also a GRUB command called "uppermem" which might help.
I doubt it. That's for ancient, broken BIOSes that don't report more
than 64k.
> (The letters SMP are talking to me. I don't know what they're saying,
> but they're talking.)
They're saying "symmetric multi-processing".
On Tue, Nov 18, 2003 at 10:47:17AM -0500, Chris wrote:
> What kernel are you using? 2.4.21 high supports starts at 4gig, older
> kernels high support is 1gig
Huh? I don't think that's true:
dice:~# cat /etc/redhat-release
Red Hat Linux release 7.3 (Valhalla)
dice:~# uname -a
Linux dice 2.4.18-19.7.xsmp #1 SMP Thu Dec 12 07:56:58 EST 2002 i686 unknown
dice:~# top -b -n 1| egrep '^CPU|^Mem'
CPU0 states: 0.0% user, 0.0% system, 0.0% nice, 100.0% idle
CPU1 states: 0.0% user, 0.1% system, 0.0% nice, 99.0% idle
Mem: 2064836K av, 2047264K used, 17572K free, 0K shrd, 154680K buff
That's a stock RedHat 7.3 kernel with some (but not the latest)
RH kernel updates applied.
Is this 1 GB limitation a symptom specific to RH 9 kernels, perhaps?
On Tue, Nov 18, 2003 at 10:55:58AM -0500, Paul wrote:
> I was going to mention that, but I'm running RH 9.0 with Red Hat's
> 2.4.20-20.9 kernel package and it is set for 4G, so I assumed this other
> RH system would be configured the same way. How can that setting be
> checked, othen than by running xconfig?
What kernel package is installed on the system that can see only 1
GB?
(You can check with `rpm -qa | grep kernel`.)
On Tue, Nov 18, 2003 at 11:14:03AM -0500, Chris wrote:
> /usr/src/linux-X-X-X/.config or the easiest make menuconfig
... which will only work if you've got kernel sources installed.
On Tue, Nov 18, 2003 at 11:16:26AM -0500, Mattison, Jacob wrote:
> I didn't set up this machine, and it does not appear to include linux
> source.
`up2date-nox -u kernel-source` should fix that, but it'll probably
be out of sync with your current kernel version unless you update
that. (There may be a way to make up2date(8) give you older kernel
versions
> I guess I need to download and compile -- in which case I might as
> well go to the newer kernel.
Right, but you might as well do that through the usual RH means if
you don't want to deal with compiling a custom kernel (and it
doesn't sound like you have any reason to).
> I've never done this before. Tell me the kernel can't smell fear..... Tell
> me it's as easy as compiling apache....
It can't and it's not.
On Tue, Nov 18, 2003 at 11:26:23AM -0500, Chris wrote:
> Be sure to copy your current kernel to a floppy incase you need to
> revert back. Upgrading a kernel is pretty easy.
Simply retaining a copy on disk and in the {grub,lilo}.conf
configuration would be sufficient.
> Once you unpack your source copy your current .config to the new source
> and do a make oldconfig unless you want to start from scratch.
He hasn't got the (RH default) .config installed, and doing make
oldconfig will just get another kernel limited to 1 GB even if he
did.
Installing RH's kernel-source package and building from that would
work, but it'd make more sense to just install the latest version of
RH's kernel package.
On Tue, Nov 18, 2003 at 11:15:37AM -0500, Chris wrote:
> I'm sure your kernel has the option, by default high memory is off.
RH turns it on by default on any version they expect to be used on
serverl-class hardware (at present, that's RH 7.2 and 7.3, RHAS 2.1).
On Tue, Nov 18, 2003 at 12:09:55PM -0500, Paul wrote:
> I know. Symetric Multi-Processing. I was thinking that maybe 1G is
> being allocated to each CPU. ?
That would be precisely NOT symmetric multi-processing, that would
be partitioned multi-processing. The point of SMP is that you don't
need to partition system resources (memory for sure, but also I/O
busses and even file systems in certain implementations) and
dedicate partitions to processors (or, if you look at it from IBM's
OS/390^H^H^H^H^H^HzOS's point of view, the other way 'round).
--
gabriel rosenkoetter
gr@eclipsed.net
Attachment:
pgpTu77ylTs2D.pgp
|
|