On 8/29/05, Carsten Otte <[EMAIL PROTECTED]> wrote:
> I would think z/VM should never put a guest on the eligible list for so long
> that people think the box is crashed. If that's the case, I'd consider it to
> be a bug.
So what options does z/VM have then when you overcommit resources and
sudde
On Aug 29, 2005, at 3:40 PM, Little, Chris wrote:
http://www.eweek.com/article2/0,1895,1853041,00.asp
"Also in the cards is an implementation of "executeI in place" for a
specific s/390 usage, kernel maintainer Andrew Morton told eWEEK."
Did I miss something? (surely not...I'm not that busy, a
Hi, Chris.
I think what Mr Morton is referring to is the "execute in place" file
system technology that IBM has made available on their developer website:
ftp://www6.software.ibm.com/software/developer/linux390/docu/l26bhe00.pdf
Execute in place is a patch that allows for a file system that hold
xip2 is a file system that appears as ext2 but resides in a DCSS when
running under z/VM. To execute the programs you execute straight from
the DCSS without having to move the data from the filesystem. You mmap
directly to the page(s) in the DCSS. This means that all your virtual
servers *can* shar
http://www.eweek.com/article2/0,1895,1853041,00.asp
"Also in the cards is an implementation of "executeI in place" for a
specific s/390 usage, kernel maintainer Andrew Morton told eWEEK."
Did I miss something? (surely not...I'm not that busy, am I?) What is
"executeI in place"? Is the "I" a typ
Thanks..
>>> [EMAIL PROTECTED] 8/29/2005 4:18:23 PM >>>
Try:
cd /var ; du -cks * | sort -rn | head -11
This will list your "culprit(s)" without having to look down the list
trying to find the largest directory. Go into the largest in the list
and do the same command again, until you
> >>> [EMAIL PROTECTED] 08/29/05 2:10 PM >>>
> David Boyes wrote:
> > Swapping to real DASD *by default* is stupid. Swapping to real DASD
> as part
> > of a planned hierarchy of swap spaces is not at all stupid.
> Since Linux is unaware of swap priorities with regard to
> "migrating hot pages to h
Try:
cd /var ; du -cks * | sort -rn | head -11
This will list your "culprit(s)" without having to look down the list trying to
find the largest directory. Go into the largest in the list and do the same
command again, until you find things you're willing to remove.
I alias this command, minus
I don't recall how this works on SLES
If you read a page from the swap disk, and you don't modify it, and it
becomes a least recently used page, will it be rewritten or just deleted
(and use the existing on the swap dataset)?
VM use to just leave the existing page alone. But people have bee
We will be using Velocity's monitor. But we need to do the migration
from the MP3000 to the z/890 first. (time constraints).
They also seem to have some sort of Oracle monitor (we will be using
Oracle a lot on z/Linux). I don't know if it is something that I can
use as an Oracle DBA or just as
If you only have one file system, then start at the root and recursively
drill down:
cd /
du -csm * | sort -n
Mark Post
-Original Message-
From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of
Mike Lovins
Sent: Monday, August 29, 2005 3:12 PM
To: LINUX-390@VM.MARIST.EDU
Subject:
cd /var
du -hs *
cd
lather, rinse, repeat until you find the culprit(s)
Mike Lovins <[EMAIL PROTECTED]>
Sent by: Linux on 390 Port
08/29/2005 03:11 PM
Please respond to Linux on 390 Port
To: LINUX-390@VM.MARIST.EDU
cc:
Subject:[LINUX-390] Disk full
On 8/29/05, Carsten Otte <[EMAIL PROTECTED]> wrote:
> Oh, I always tend to measure on my little poor VM guest on this development
> Trex. And because there are other users, my measurements are never very
> acurate and therefore I do not post them. This is why I asked Barton if he
> actually meaure
David Boyes wrote:
> Swapping to real DASD *by default* is stupid. Swapping to real DASD as part
> of a planned hierarchy of swap spaces is not at all stupid.
Since Linux is unaware of swap priorities with regard to "migrating hot pages
to high prio swap", as soon as you have part of your hot worki
I have noticed that my system disk is full. I look in /var/log and found
numerious mail.x.gz, warn.xx.gz nad messages.xx.gz files. I
removed all of these files. I also noticed that the mail and warn files
were very large. I ran logrotate and it reduced the file sizes and
greated more .g
Adam Thornton wrote:
> I don't know whether stock Debian Sarge will support it, but when I
> get a chance I can find out and build some kernels that will if the
> base one doesn't.
I looked at the (vanilla) 2.6.8.2 kernel and at the ibm.diff in the
kernel pagage. That does'nt seem like it would run
> For my setup it makes a significant difference in favour of
> dcss (more than factor two faster than vdisk, even more then
> that in cpu horsepower saving). And I really have'nt seen a
> measurement showing any advantage of vdisk so far.
1) It works with all releases of Linux.
2) It is substant
Adam Thornton wrote:
> Do you have performance data for this?
Oh, I always tend to measure on my little poor VM guest on this development
Trex. And because there are other users, my measurements are never very
acurate and therefore I do not post them. This is why I asked Barton if he
actually meaur
Sorry, put the files in the wrong order when I did the diff for the
oorexx.spec file. This is the correct diff:
linuxsjs:/usr/src/packages/SPECS # cat oorexx.spec.diff
--- oorexx.spec.orig2005-08-29 10:00:21.108911753 -0400
+++ oorexx.spec 2005-08-29 10:01:58.0 -0400
@@ -65,7 +65,7 @
Tom Duerbusch wrote:
> "IF" you have sufficient resources, that I agree that swapping to real
> disk is stupid. But I have done a lot of swapping to real disk because
> of real memory constraints.
But z/VM would also put vdisk contents to its paging, and z/VM would self-
evident page to real dasd.
I was able to get this successfully installed as a 32-bit application.
This is what I did:
1) On my sandbox 64-bit system, install using YaST the following packages:
binutils-32bit
glibc-32bit
glibc-devel-32bit
glibc-local-32bit
libgcc-32bit
ncurses-32bit
ncurses-devel-32bit
2) Download and i
In your procedure,
fdasd /dev/dasdxx-- add a Linux partition to it
p
n (take defaults by pressing enter twice for starting and ending cyls)
p
W
Can be changed to
fdasd -a /dev/dasdxx
Which will create one partition on the disk, with no questions asked.
--
Robert P. Nix Mayo F
I forwarded the thread regarding ooRexx to Chip Davis, President of the
Rexx Language Association, which has responsibility for ooRexx development
and enhancement now. The following is from his reply to me:
We're currently working on a 64-bit version of ooRexx but
at the moment all that's availab
We have been doing this for more than a year. The server in on Linux
SuSE8, and clinet on USS os/390. I usually code in Linux, and ftp to
USS. The only change I make, is to change g++ to cxx in USS Makefile.
All else is the same.
Richar Stevense's Unix Network programmin valume 1 is a good pla
On Aug 29, 2005, at 4:11 AM, Carsten Otte wrote:
Post, Mark K wrote:
Barton Robinson once told me he thought
VDISK could sustain a much higher paging rate than I would have
thought
reasonable, and not impact performance noticeably.
Swap to dcssblk works even better, but you need Sles9 >= SP2
"IF" you have sufficient resources, that I agree that swapping to real
disk is stupid. But I have done a lot of swapping to real disk because
of real memory constraints.
I've had a MP3000 with 1 GB of memory, running 9 VSEs and 17 Linux
guests. Most of the time, VM paging is low, 5-15 per second
There have been any number of people over the last 5 years that have
seen just that happen. It's generated enough mail on this list that
asking how "big" the guest is has become one of the first questions
asked. It's z/VM "working as designed."
Mark Post
-Original Message-
From: Linux
Barton Robinson wrote:
> When doing this, it is important to use vdisk and enough
> of it. AND have vdisk for swap implemented correctly.
> I saw a presentation at share that said to use dasd and
> not vdisk for swap. ignore this advice if you get it.
> people not doing performance research making
Now that SLES9 has the I/O fixed buffer, that has also been
retrofitted to SLES8 as of a couple weeks ago, there is more
of an opportunity to push oracle performance limits. On the
redbook draft just published for oracle, i was hoping to
work on pushing the storage performance limits,
but the i/o l
On Monday, 08/29/2005 at 12:44 ZE2, Carsten Otte <[EMAIL PROTECTED]> wrote:
> I would think z/VM should never put a guest on the eligible list for so
long
> that people think the box is crashed. If that's the case, I'd consider
it to
> be a bug.
I don't think I'd call it a bug. Rather, the system
I'm going to add some more info to my CSE presentation before I send it to
you Mark, but it's coming...
--
Jay Brenneman
--
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] wit
Little, Chris wrote:
> An extension of this . . .
>
> Will shared memory swap?
Yes, same as non-shared.
--
Carsten Otte
IBM Linux technology center
ARCH=s390
--
For LINUX-390 subscribe / signoff / archive access instructions,
se
Post, Mark K wrote:
> Barton Robinson once told me he thought
> VDISK could sustain a much higher paging rate than I would have thought
> reasonable, and not impact performance noticeably.
Swap to dcssblk works even better, but you need Sles9 >= SP2 for that.
--
Carsten Otte
IBM Linux technology
Tom Duerbusch wrote:
> Same problems occur over on the tape side. The IBM 3590s and above can
> be channel attached (ficon) or fiber attached (fcp). And then you have
> to divide your drives into ficon or fcp attached...no sharing. Increase
> in number of tape drives, increase in the number of c
Tom Duerbusch wrote:
> 1. Take native scsi, and with microcode, emulate mainframe dasd (CKD
> or FBA), which takes controller cycles.
> 2. Ship it over a FICON channel to the host.
> 3. z/LINUX has a piece of code that emulates scsi on mainframe DASD
> (which takes mainframe CPU cycles).
No. The
Post, Mark K wrote:
> That is true if/when the Linux system finally runs out of virtual
> storage and swap space. If it just causes the instance's working set
> size to get very large, z/VM might decide to put it on the eligible list
> for a long time.
I would think z/VM should never put a guest o
36 matches
Mail list logo