ci-linux-devel — Discussion about Cluster Infrastructure for Linux

[CI] On holiday...

[Brian J. W. <Bri...@co...>]

..in Europe for a couple of weeks. I'm leaving Friday, Oct. 12, and
will be back Monday, Oct. 29. I won't be checking e-mail during this
time.

-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

[CI] Re: [SSI] Back online (Firewire)

[Brian J. W. <Bri...@co...>]

Lyle Bickley wrote:
> 
> I checked the GFS (Sistina), and IEEE 1394 FAQ's and mail lists and
> found that firewire is specified to be capable of sharing a HDD with
> multiple systems, but that no vendors are currently supporting that
> capability in hardware.
> 
> Those who have tested a single firewire HDD with multiple system access
> have said that whichever system initially gains control over a firewire
> HDD, it is never released - and therefore the HDD cannot be accessed by
> a second system.
> 
> As I recollect, there were suggestions that Microsoft didn't want to
> support firewire HDD sharing in Windows, therefore the HW manf. didn't
> bother to implement the capability (not verified by me - but it does
> kinda ring true).
> 

Hmmm. Well I guess that moots the argument for now.

-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

[CI] Re: [SSI] Back online (Firewire)

[Lyle B. <lbi...@bi...>]

I checked the GFS (Sistina), and IEEE 1394 FAQ's and mail lists and
found that firewire is specified to be capable of sharing a HDD with
multiple systems, but that no vendors are currently supporting that
capability in hardware.

Those who have tested a single firewire HDD with multiple system access
have said that whichever system initially gains control over a firewire
HDD, it is never released - and therefore the HDD cannot be accessed by
a second system.

As I recollect, there were suggestions that Microsoft didn't want to
support firewire HDD sharing in Windows, therefore the HW manf. didn't
bother to implement the capability (not verified by me - but it does
kinda ring true).

Lyle

"Brian J. Watson" wrote:
> 
> Christoph Hellwig wrote:
> > > I'm not sure how difficult of a hack it would be to get GFS working on
> > > FireWire. You might pose this to Sistina and the OpenGFS group.
> >
> > It will run over ieee1394 surely - the prolem is lack of memexp support.
> 
> Isn't this just a problem for DMEP hardware on IEEE1394? My
> understanding of the memexpd form of locking is that it is strictly IP
> based.
> 
> --
> Brian Watson                | "The common people of England... so
> Linux Kernel Developer      |  jealous of their liberty, but like the
> Open SSI Clustering Project |  common people of most other countries
> Compaq Computer Corp        |  never rightly considering wherein it
> Los Angeles, CA             |  consists..."
>                             |     -Adam Smith, Wealth of Nations, 1776
> 
> mailto:Bri...@co...
> http://opensource.compaq.com/

-- 
Lyle Bickley             | Bickley Consulting West Inc.
lbi...@ac...         | 
lbi...@bi... | V 650-428-0621
http://bickleywest.com/	 | F 650-428-0599
		
"Black holes exist where GOD is dividing by zero"

[CI] Re: [SSI] Back online

[Brian J. W. <Bri...@co...>]

Christoph Hellwig wrote:
> > I'm not sure how difficult of a hack it would be to get GFS working on
> > FireWire. You might pose this to Sistina and the OpenGFS group.
> 
> It will run over ieee1394 surely - the prolem is lack of memexp support.

Isn't this just a problem for DMEP hardware on IEEE1394? My
understanding of the memexpd form of locking is that it is strictly IP
based.


-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

[CI] Re: [SSI] Back online

[Christoph H. <hc...@ns...>]

On Tue, Sep 18, 2001 at 06:24:10PM -0700, Brian J. Watson wrote:
> 
> > (It's a shame that the firewire drivers don't support GFS - that would
> > be a lot cheaper than FC...)
> > 
> 
> I'm not sure how difficult of a hack it would be to get GFS working on
> FireWire. You might pose this to Sistina and the OpenGFS group.

It will run over ieee1394 surely - the prolem is lack of memexp support.
Maybe there are similar ieee1394 commands and we could do a translation
in the driverbut I doubt it.

That leaves IP-based locking..

	Christoph

-- 
Of course it doesn't work. We've performed a software upgrade.

[CI] Re: [SSI] Back online

[Brian J. W. <Bri...@co...>]

Lyle Bickley wrote:
> 
> I sent this message previously while VA Linux was moving things around -
> and it likely got lost...
> 
> Bruce, Brian, et al,
> 
> It was good chatting with you in person at Linuxworld!  Your demo there
> encouraged me to setup a SSI cluster in my lab.  I have a small beowulf
> cluster (8 nodes) - 4 of which are Intel Uni nodes.  It looks like I'll
> have to start with a 2 node SSI cluster :(  as I will have to use a SCSI
> external disk on the first cut for GFS.
> 

Cool! Nice to meet you, as well. I'm glad you're interested in giving
SSI a spin.


> You mentioned that you had corruption troubles with one drive type when
> you attempted to use SCSI clustering.  Were you using SCSI2 or SCSI3?
> What kind of drive was it?
> 

Dave could answer this better than I.


> (It's a shame that the firewire drivers don't support GFS - that would
> be a lot cheaper than FC...)
> 

I'm not sure how difficult of a hack it would be to get GFS working on
FireWire. You might pose this to Sistina and the OpenGFS group.


> Are there any special considerations for SSI regarding the size of swap
> I need to know?
> 

I don't think so. John?


> The documented instructions generally seem clear enough  - but what
> would help a lot would be a diagram of the your demo system layout
> (names, IP addresses, etc.) along with the config files you used to
> implement it.  That would probably answer a number of minor questions I
> have regarding CLUSTER
> params, etc.
> 

See the attached image for a rough diagram. The default config files
installed as a part of step III.1 of the INSTALL file should be mostly
correct. One problem I spotted is that the lockdev field of
/etc/gfscf.cf should be an IP address on the same network as the cluster
interconnect (such as 192.168.1.5).

Just follow the installation instructions carefully. Our demo system was
installed by someone in Houston (we're in L.A.) unfamiliar with what
we've done on Linux. He mostly did it by following the instructions in
the INSTALL file, although we had to work with him a bit over the phone
to deal with bugs he discovered. We tried to use this experience to
improve the INSTALL file in the 0.5.1 release of SSI.

If you run into specific problems, please let us know.


-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

Re: [CI] Support for old kernels

[Brian J. W. <Bri...@co...>]

Marc-Christian Petersen wrote:
> 
> Hi Brian,
> 
> >We would like to get some feedback on which kernel versions we should
> >release CI patches against. Right now we have patches for vanilla
> >2.4.2 and 2.4.4. Very soon, we're planning to release CI 0.5.6, which
> >will support Alpha (in addition to x86) and include some bug fixes.
> I think a patch for the 2.2.19 Kernel could be very interesting, cause
> so
> many ppl (including me) are using 2.2.19 cause that was the real last
> stable
> kernel. And if you would do some actual patches for the 2.4 Kernel, i
> think
> it should be for 2.4.9-ac10 or later.


We don't have the resources to back-port our stuff to the 2.2 series. It
would be a non-trivial task for SSI, considering the number of hooks we
have in the base that would need to be reevaluated. There might also be
serious deadlock problems due to the lack of fine-grain locking in 2.2.
As it is, we've had to be a bit creative in 2.4. It's best to keep SSI
moving forward in kernel versions.

For CI, it might be a bit easier. CLMS (the membership subsystem) is
self-contained, without dependencies on the base. ICS (internode
communication) is already divided into transport dependent and
independent code. The transport dependent routines follow the ics_ll*
naming convention. They make use of the base TCP structures and
functions, and would need to be modified accordingly. There are also a
few hooks in the TCP code itself that would need to be adapted. We have
no plans to do this backport, but I don't think it would be terribly
difficult for someone else to do.

What's the advantage of being on Alan's kernel versus Linus'?


> 
> >For that release, we would like to bring the patch up to 2.4.6 and
> >just release for that kernel version. Are there any objections?
> The only objection i have is that the kernel 2.4.6 is very old and was
> not so
> stable as 2.4.4. Some of my friends and me had many troubles with the
> 2.4.6
> Kernel.
> 

I know. I think we moved up to 2.4.6 before it was out for very long,
then we got caught up in cranking out the SSI release, and preparing for
and doing a demo at Caldera Forum and LinuxWorld. We're now preparing
another SSI release with some fixes for bugs and memory leaks, and we
think we'll merge it up to 2.4.9 or 2.4.10. CI, being a subset of SSI,
should also merge up. I'll send a separate, shorter message to these
lists asking for opinions about this.


> >Once 2.4 has stabilized and been handed off to Alan Cox, it may make
> sense
> >to support more than one kernel version. At that point, people might be
> 
> >using old kernels because other patches they need haven't been updated
> in
> >awhile. Until then, however, I'm guessing that anybody who knows how to
> 
> >build a kernel wouldn't mind being on one of the latest and (hopefully)
> 
> >least buggy bases. Please let us know what you think.
> I think, 2.4.7 or 2.4.4 are more stable than 2.4.6, even the 2.4.9acX is
> very
> stable and therefore i think a newer CI Patch for a newer kernel would
> be
> very nice. I agree, 2.4 has not been fully stableized, but 2.4.6 is not
> as
> stable as ... what i write?! ... i wrote it above :-)


Keep in mind that both CI and SSI are pre-alpha. The instabilities in
2.4.6 may not be significantly greater than any problems lurking in our
own code. Of course, we would prefer to be on a stable base. ;)


> 
> Let me know what you think about my comments! :-)
> 

Thanks for the feedback!


> --
> 
> Kind regards
>         Marc-Christian Petersen
> 
> Linux-Systeme GmbH
> Hinseler Hof 3a, 45277 Essen
> Tel.: +49 201 - 85 85 130
> Fax : +49 201 - 75 00 1579
> D2  : +49 173 - 541 68 09
> mailto: mc...@li...
> http://www.linux-systeme.de/

-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

Re: [CI] Including kdb

[<jd...@al...>]

On Mon, Sep 10, 2001 at 05:11:12PM -0700, Brian J. Watson wrote:
> We're thinking of rolling our kdb into the next releases of CI and SSI,
> to help with bug reports. At the very least, it's nice to know what the
> stack trace was leading up to a NULL or bad pointer dereference.

I know John is real busy, but do you guys think he'd have time to
contribute the 'call' modification  as a patch back to the main kdb
tree?  It would be a great tool to have for us regular Linux kernel
hackers, not just you cool SSI guys...

Jason Duell
Lawrence Berkeley Laboratory

[CI] Including kdb

[Brian J. W. <Bri...@co...>]

We're thinking of rolling our kdb into the next releases of CI and SSI,
to help with bug reports. At the very least, it's nice to know what the
stack trace was leading up to a NULL or bad pointer dereference.

Our kdb is SGI's kdb, modified to patch cleanly against a SSI kernel,
and enhanced with a call command for calling kernel functions such as
debug print routines. For those who are not comfortable poking around
the innards of a kernel, we would ask that if something goes wrong and a
kdb> prompt pops up, just type bt and hit Enter. If you're running over
a serial console, you could then copy and paste the resulting stack
trace into a mail message. If you're on a physical console, retyping the
whole thing is a bit of a drag. We would just ask that you copy a subset
of the information into a mail message.

Are there any comments, objections, or questions about this?

-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

[CI] Re: [SSI] Back online

[Lyle B. <lbi...@bi...>]

I sent this message previously while VA Linux was moving things around -
and it likely got lost...

Bruce, Brian, et al,

It was good chatting with you in person at Linuxworld!  Your demo there
encouraged me to setup a SSI cluster in my lab.  I have a small beowulf
cluster (8 nodes) - 4 of which are Intel Uni nodes.  It looks like I'll
have to start with a 2 node SSI cluster :(  as I will have to use a SCSI
external disk on the first cut for GFS.

You mentioned that you had corruption troubles with one drive type when
you attempted to use SCSI clustering.  Were you using SCSI2 or SCSI3? 
What kind of drive was it?

(It's a shame that the firewire drivers don't support GFS - that would
be a lot cheaper than FC...)

Are there any special considerations for SSI regarding the size of swap
I need to know?

The documented instructions generally seem clear enough  - but what
would help a lot would be a diagram of the your demo system layout
(names, IP addresses, etc.) along with the config files you used to
implement it.  That would probably answer a number of minor questions I
have regarding CLUSTER
params, etc.

I won't start working on this until mid-late September - but before I
start, if you have you could supply the demo info. it would be greatly
appreciated.

I'll keep you posted as to my progress.

Thanks,
Lyle

-- 
Lyle Bickley             | Bickley Consulting West Inc.
lbi...@ac...         | 
lbi...@bi... | V 650-428-0621
http://bickleywest.com/	 | F 650-428-0599
		
"Black holes exist where GOD is dividing by zero"

[CI] Back online

[Brian J. W. <Bri...@co...>]

Kitrick Sheets wrote:
> 
> Welcome back. Been wondering where you went 8^).
> 
> Kitrick
> 

Thanks. While Bruce and I were up in the San Francisco Bay Area, touting
the glory of SSI to all who cared to listen, VA Linux was busy moving
our stuff over to their servers. Unfortunately the move was less than
transparent for our mailing lists. As you can see, our mailing lists are
just now back online.

Note that the address to post messages has changed from the old
'@opensource.compaq.com' variety to:
  ssi...@li...
  ci-...@li...

Also note that any messages sent between Aug 22 and Sept 7 may have
fallen into a black hole as far as the archiver is concerned.
Previously, we had used Pipermail on a Compaq server. It stopped
receiving e-mails on Aug 22, when the opensource.compaq.com DNS entry
was switched over to a SourceForge machine. On Sept 7, when VA Linux
unstuck our mailing lists, they also added a Geocrawler agent as a
member of both lists.

The upshot is twofold. First, there are two mail archives per project,
accessible from our project homepages. The first archive is the current
Geocrawler archive, and the second is a snapshot of the old Pipermail
archive. The second upshot is that if you posted any messages that
didn't make it into either archive, and you think it should, please
resend it with any appropriate context.


-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

[CI] test

[Marc M. <me...@va...>]

testing list and archiving

Marc
-- 
VA Linux Systems Server Sysadmin / Sourceforge mail&list master. 510 687 7061
 
Home page: http://marc.merlins.org/  - Finger ma...@me... for PGP key
1024R/763BE901 / Fingerprint = A1 9F 94 B7 78 01 E5 21  21 E0 F1 2E A2 85 E2 77

[CI] CI Testing

[Brian J. W. <Bri...@co...>]

Just making sure everything works now that we've migrated to
SourceForge.

-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

[CI] test. plz ignore

[John H. <sn...@va...>]

Test message ...

-- 
---------------------------------------------------------------------
      / \            John Hoffmann                   Systems Engineer
     / . \           SourceForge OnSite          VA LiNUX Systems Inc
\ . /                http://www.valinux.com        sn...@va...
 \_/  Linux Systems  Work: (510) 687-6879        Cell: (613) 290-5217
                           SMS: jho...@tx...
---------------------------------------------------------------------
This is Linux Country. On a quiet night, you can hear Windows reboot.

[CI] Testing #2

[Brian J. W. <Bri...@co...>]

Just making sure everything works now that we've migrated to
SourceForge.

-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/

Re: [CI] Thus this fit our requirment documentation

[Lyle B. <lbi...@bi...>]

Peter,

The new link works fine outside of the IBM firewall.

Thanks,
Lyle

Peter Badovinatz wrote:
> 
> --- Lyle Bickley <lbi...@bi...> wrote:
> > Peter,
> >
> > The link:  http://www.sun.com/clusters/dh.brown.pdf
> >
> > Appears to be out-of-date.  Do know of a copy elsewhere?
> 
> Lyle,
> 
> Try:  http://www.rs6000.ibm.com/resource/consult/dhbrown/rasc.html
> 
> I believe this should be accessible from outside the IBM firewall.  I don't
> have immediate access to a machine "outside" so I can't test.  Please let me
> know if you can get to this.  If so, I'll ask Alan to update the link on
> linux-ha.org.
> 
> Peter
> 
> > Cheers,
> > Lyle
> >
> > Peter Badovinatz wrote:
> > >
> > > --- Aneesh Kumar <kva...@ya...> wrote:
> > > > Hi ,
> > > >
> > > >  Found this document. I gives a detailed analysis of
> > > > what actually one look in a cluster.  Can we see Linux
> > > > listing in some of these studies in another  2-3
> > > > years.
> > > >
> > > >
> > > > http://tru64unix.compaq.com/dhba_ras.pdf
> > >
> > > Take a look at http://www.linux-ha.org/.  This document, and the same
> > report
> > > from 1998 (http://www.sun.com/clusters/dh.brown.pdf), both by D.H. Brown,
> > are
> > > both referenced there.  You'll also see a rather over-optimistic statement
> > > there in the paragraph:
> <snip>
> > >
> > > Peter
> > >
> > > =====
> > > These have been the opinions of:
> > > Peter R. Badovinatz -- (503)578-5530 (TL 775)
> > > wo...@us.../tab...@ya...
> > > and in no way should be construed as official opinion of
> > > IBM, Corp.
> >
> > --
> > Lyle Bickley
> > Bickley Consulting West Inc.
> > lbi...@bi... | VM 650-428-0621
> > http://bickleywest.com   | FX 650-428-0599
> 
> =====
> These have been the opinions of:
> Peter R. Badovinatz -- (503)578-5530 (TL 775)
> wo...@us.../tab...@ya...
> and in no way should be construed as official opinion of
> IBM, Corp.
> 
> __________________________________________________
> Do You Yahoo!?
> Make international calls for as low as $.04/minute with Yahoo! Messenger
> http://phonecard.yahoo.com/

-- 
Lyle Bickley
Bickley Consulting West Inc.
lbi...@bi... | VM 650-428-0621
http://bickleywest.com   | FX 650-428-0599

[CI] IBM DLM information

[Peter B. <tab...@ya...>]

--- "Aneesh Kumar K.V" <ane...@di...> wrote:
> Hi , 
> 
>  Can you just send me  the link for "programming locking application ". The 
> link that DLM page have is taking me to HA documentation. It doesn't talk
> about  locking APIs.
> 
>  -aneesh 
> 
Aneesh,

Which page are you looking at?  Let me know and if there's a problem with a
link I'll get it fixed.

If you look at our current DLM pages, at:
http://oss.software.ibm.com/developerworks/projects/dlm
or
http://oss.software.ibm.com/developerworks/opensource/dlm/

Both point to the new Linux-specific DLM client programming guide (look for
references to "Review draft"), it is available as html and as pdf.  This is an
adaptation of the 'Programming Locking Applications' book from HACMP on AIX. 
This book will describe using the DLM as a client, although it does not do much
to describe the internals of the DLM.

The structure of the DLM is such that it consists of core code that is loaded
into the kernel which provides the specific locking functionality.  The code
that interacts with clustering code (e.g., heartbeat, Compaq CI, etc.) is in
the cvs tree in source/dlmdu and source/dlmdcl.  The CVS tree is accessible
from http://oss.software.ibm.com/developerworks/projects/dlm or you can simply
download the most recent 'release' from that page as well.

What information exists about the DLM internals is in CVS in the documents
directory.

Peter


=====
These have been the opinions of:
Peter R. Badovinatz -- (503)578-5530 (TL 775)
wo...@us.../tab...@ya...
and in no way should be construed as official opinion of 
IBM, Corp.

__________________________________________________
Do You Yahoo!?
Make international calls for as low as $.04/minute with Yahoo! Messenger
http://phonecard.yahoo.com/

Re: [CI] Thus this fit our requirment documentation

[Peter B. <tab...@ya...>]

--- Lyle Bickley <lbi...@bi...> wrote:
> Peter,
> 
> The link:  http://www.sun.com/clusters/dh.brown.pdf
> 
> Appears to be out-of-date.  Do know of a copy elsewhere?

Lyle,

Try:  http://www.rs6000.ibm.com/resource/consult/dhbrown/rasc.html

I believe this should be accessible from outside the IBM firewall.  I don't
have immediate access to a machine "outside" so I can't test.  Please let me
know if you can get to this.  If so, I'll ask Alan to update the link on
linux-ha.org.

Peter

> Cheers,
> Lyle
> 
> Peter Badovinatz wrote:
> > 
> > --- Aneesh Kumar <kva...@ya...> wrote:
> > > Hi ,
> > >
> > >  Found this document. I gives a detailed analysis of
> > > what actually one look in a cluster.  Can we see Linux
> > > listing in some of these studies in another  2-3
> > > years.
> > >
> > >
> > > http://tru64unix.compaq.com/dhba_ras.pdf
> > 
> > Take a look at http://www.linux-ha.org/.  This document, and the same
> report
> > from 1998 (http://www.sun.com/clusters/dh.brown.pdf), both by D.H. Brown,
> are
> > both referenced there.  You'll also see a rather over-optimistic statement
> > there in the paragraph:
<snip>
> > 
> > Peter
> > 
> > =====
> > These have been the opinions of:
> > Peter R. Badovinatz -- (503)578-5530 (TL 775)
> > wo...@us.../tab...@ya...
> > and in no way should be construed as official opinion of
> > IBM, Corp.
> 
> -- 
> Lyle Bickley
> Bickley Consulting West Inc.
> lbi...@bi... | VM 650-428-0621
> http://bickleywest.com   | FX 650-428-0599


=====
These have been the opinions of:
Peter R. Badovinatz -- (503)578-5530 (TL 775)
wo...@us.../tab...@ya...
and in no way should be construed as official opinion of 
IBM, Corp.

__________________________________________________
Do You Yahoo!?
Make international calls for as low as $.04/minute with Yahoo! Messenger
http://phonecard.yahoo.com/

[CI] (no subject)

[<kva...@ya...>]

Hi,
Do you have any documentation that discuss about the
architecture of the DLM. Suppose if i want to have it
on TOP of CI-Linux which part I should paly with? 

-aneesh 

On 07 Aug 2001 00:06:09 -0700, Peter Badovinatz wrote:
> 
> --- Aneesh Kumar <kva...@ya...> wrote:
> > Hi , 
> > 
> >  Is there any one working on IBM's DLM on 
CI-Linux ?
> > 
> 
> Yes and no, so to speak.  We intend to work with
CI-Linux, we have not done
> much with it yet, due to us still working on the DLM
porting to Linux in
> general.  The need, as we see it, is to create a
'glue' module that allows the
> DLM core to get membership events from CI-Linux. 
There should be no internal
> changes necessary to the core DLM functionality. 
This has been an intentional
> aim of our DLM.
> 
> I have no specific schedule I can provide at this
time about when we'll release
> anything, if it's up to us it will have to wait for
a few weeks until we can
> get to it.
> 
> Patches gratefully accepted.
> 
> >  -aneesh 
> > 
> 
> Peter
> 
> =====
> These have been the opinions of:
> Peter R. Badovinatz -- (503)578-5530 (TL 775)
> wo...@us.../tab...@ya...
> and in no way should be construed as official
opinion of 
> IBM, Corp.
> 
>


____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie

Re: [CI] Thus this fit our requirment documentation

[Lyle B. <lbi...@bi...>]

Peter,

The link:  http://www.sun.com/clusters/dh.brown.pdf

Appears to be out-of-date.  Do know of a copy elsewhere?

Cheers,
Lyle

Peter Badovinatz wrote:
> 
> --- Aneesh Kumar <kva...@ya...> wrote:
> > Hi ,
> >
> >  Found this document. I gives a detailed analysis of
> > what actually one look in a cluster.  Can we see Linux
> > listing in some of these studies in another  2-3
> > years.
> >
> >
> > http://tru64unix.compaq.com/dhba_ras.pdf
> 
> Take a look at http://www.linux-ha.org/.  This document, and the same report
> from 1998 (http://www.sun.com/clusters/dh.brown.pdf), both by D.H. Brown, are
> both referenced there.  You'll also see a rather over-optimistic statement
> there in the paragraph:
> 
> <begin quote>
> "We are on the road to being competitive with commercial systems similar to
> those described in D. H. Brown's 1998 or March 2000 analysis of RAS cluster
> features and functions.  As noted below, we have added a new project to port
> SGI's FailSafe product to Linux. This will help accelerate the process
> significantly.  It seems possible for Linux HA software to achieve D. H. Brown
> level of competitiveness in 2001."
> <end quote>
> 
> Are we there as of 2001?  IMHO, no.  But if you read linux-ha.org and follow
> the various links, especially discussions related to directions and frameworks,
> then we may see our way to making it to that future.
> 
> >
> >  -aneesh
> 
> Peter
> 
> =====
> These have been the opinions of:
> Peter R. Badovinatz -- (503)578-5530 (TL 775)
> wo...@us.../tab...@ya...
> and in no way should be construed as official opinion of
> IBM, Corp.
> 
> __________________________________________________
> Do You Yahoo!?
> Make international calls for as low as $.04/minute with Yahoo! Messenger
> http://phonecard.yahoo.com/
> _______________________________________________
> ci-linux-devel mailing list
> ci-...@op...
> http://www.opensource.compaq.com/mailman/listinfo/ci-linux-devel

-- 
Lyle Bickley
Bickley Consulting West Inc.
lbi...@bi... | VM 650-428-0621
http://bickleywest.com   | FX 650-428-0599

Re: [CI] Thus this fit our requirment documentation

[Peter B. <tab...@ya...>]

--- Aneesh Kumar <kva...@ya...> wrote:
> Hi , 
> 
>  Found this document. I gives a detailed analysis of
> what actually one look in a cluster.  Can we see Linux
> listing in some of these studies in another  2-3
> years. 
> 
> 
> http://tru64unix.compaq.com/dhba_ras.pdf

Take a look at http://www.linux-ha.org/.  This document, and the same report
from 1998 (http://www.sun.com/clusters/dh.brown.pdf), both by D.H. Brown, are
both referenced there.  You'll also see a rather over-optimistic statement
there in the paragraph:

<begin quote>
"We are on the road to being competitive with commercial systems similar to
those described in D. H. Brown's 1998 or March 2000 analysis of RAS cluster
features and functions.  As noted below, we have added a new project to port
SGI's FailSafe product to Linux. This will help accelerate the process
significantly.  It seems possible for Linux HA software to achieve D. H. Brown
level of competitiveness in 2001."
<end quote>

Are we there as of 2001?  IMHO, no.  But if you read linux-ha.org and follow
the various links, especially discussions related to directions and frameworks,
then we may see our way to making it to that future.

> 
>  -aneesh 

Peter

=====
These have been the opinions of:
Peter R. Badovinatz -- (503)578-5530 (TL 775)
wo...@us.../tab...@ya...
and in no way should be construed as official opinion of 
IBM, Corp.

__________________________________________________
Do You Yahoo!?
Make international calls for as low as $.04/minute with Yahoo! Messenger
http://phonecard.yahoo.com/

Re: [CI] Any one working on DLM on CIL

[Peter B. <tab...@ya...>]

--- Aneesh Kumar <kva...@ya...> wrote:
> Hi , 
> 
>  Is there any one working on IBM's DLM on  CI-Linux ?
> 

Yes and no, so to speak.  We intend to work with CI-Linux, we have not done
much with it yet, due to us still working on the DLM porting to Linux in
general.  The need, as we see it, is to create a 'glue' module that allows the
DLM core to get membership events from CI-Linux.  There should be no internal
changes necessary to the core DLM functionality.  This has been an intentional
aim of our DLM.

I have no specific schedule I can provide at this time about when we'll release
anything, if it's up to us it will have to wait for a few weeks until we can
get to it.

Patches gratefully accepted.

>  -aneesh 
> 

Peter

=====
These have been the opinions of:
Peter R. Badovinatz -- (503)578-5530 (TL 775)
wo...@us.../tab...@ya...
and in no way should be construed as official opinion of 
IBM, Corp.

__________________________________________________
Do You Yahoo!?
Make international calls for as low as $.04/minute with Yahoo! Messenger
http://phonecard.yahoo.com/

[CI] Any one working on DLM on CIL

[<kva...@ya...>]

Hi , 

 Is there any one working on IBM's DLM on  CI-Linux ?

 -aneesh 

____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie

[CI] Thus this fit our requirment documentation

[<kva...@ya...>]

Hi , 

 Found this document. I gives a detailed analysis of
what actually one look in a cluster.  Can we see Linux
listing in some of these studies in another  2-3
years. 


http://tru64unix.compaq.com/dhba_ras.pdf

 -aneesh 


____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie

[CI] Testing...

[Brian J. W. <Bri...@co...>]

-- 
Brian Watson                | "The common people of England... so 
Linux Kernel Developer      |  jealous of their liberty, but like the 
Open SSI Clustering Project |  common people of most other countries 
Compaq Computer Corp        |  never rightly considering wherein it 
Los Angeles, CA             |  consists..."
                            |     -Adam Smith, Wealth of Nations, 1776

mailto:Bri...@co...
http://opensource.compaq.com/