From agshew at gmail.com  Mon Jun  1 10:39:42 2009
From: agshew at gmail.com (Andrew Shewmaker)
Date: Mon, 1 Jun 2009 11:39:42 -0600
Subject: [Warewulf] debugging provisiond
Message-ID: <f4050c9e0906011039k30776809g48bab48507039782@mail.gmail.com>

I'm having an issue with the passwdfile module where it isn't putting
the contents of /etc/perceus/modules/passwdfile/all into /etc/passwd

I've double checked that the module is activated, and the link to the
nodescript is in place.  I've also made sure that /etc/passwd is
writeable.

I've restarted the perceus service, and I've recreated my hybrid image.

I've passed in the enable-debug=2 kernel parameter in, but I don't see
anything useful on the node's console while it is booting.

What should I be looking at?  I must be doing something dumb.

Thank you.

-- 
Andrew Shewmaker


From agshew at gmail.com  Mon Jun  1 12:53:38 2009
From: agshew at gmail.com (Andrew Shewmaker)
Date: Mon, 1 Jun 2009 13:53:38 -0600
Subject: [Warewulf] debugging provisiond
In-Reply-To: <f4050c9e0906011039k30776809g48bab48507039782@mail.gmail.com>
References: <f4050c9e0906011039k30776809g48bab48507039782@mail.gmail.com>
Message-ID: <f4050c9e0906011253l6d4d7631r52ca40e2b2d3099a@mail.gmail.com>

On Mon, Jun 1, 2009 at 11:39 AM, Andrew Shewmaker <agshew at gmail.com> wrote:
> I'm having an issue with the passwdfile module where it isn't putting
> the contents of /etc/perceus/modules/passwdfile/all into /etc/passwd
...
> What should I be looking at?  I must be doing something dumb.

Like having perceus-provisiond installed in the vnfs, but with its
files listed in the hybridize file.  I didn't get symlinks for
provisiond, but I did for other stuff.

I think I have it now.

-- 
Andrew Shewmaker


From gmkurtzer at gmail.com  Mon Jun  1 13:42:36 2009
From: gmkurtzer at gmail.com (Greg Kurtzer)
Date: Mon, 1 Jun 2009 13:42:36 -0700
Subject: [Warewulf] debugging provisiond
In-Reply-To: <f4050c9e0906011253l6d4d7631r52ca40e2b2d3099a@mail.gmail.com>
References: <f4050c9e0906011039k30776809g48bab48507039782@mail.gmail.com>
	<f4050c9e0906011253l6d4d7631r52ca40e2b2d3099a@mail.gmail.com>
Message-ID: <571f1a060906011342h20988a0bo27e9cbef6664c04b@mail.gmail.com>

Hello Andrew,

Good catch... Now I wonder why!?

If you notice anything further please let us know. Otherwise, please
send the hybridized entries for provisiond so we can try and
replicate.

Thanks,
Greg

On Mon, Jun 1, 2009 at 12:53 PM, Andrew Shewmaker <agshew at gmail.com> wrote:
> On Mon, Jun 1, 2009 at 11:39 AM, Andrew Shewmaker <agshew at gmail.com> wrote:
>> I'm having an issue with the passwdfile module where it isn't putting
>> the contents of /etc/perceus/modules/passwdfile/all into /etc/passwd
> ...
>> What should I be looking at? ?I must be doing something dumb.
>
> Like having perceus-provisiond installed in the vnfs, but with its
> files listed in the hybridize file. ?I didn't get symlinks for
> provisiond, but I did for other stuff.
>
> I think I have it now.
>
> --
> Andrew Shewmaker
> _______________________________________________
> Warewulf mailing list
> Warewulf at caoslinux.org
> http://lists.caosity.org/mailman/listinfo/warewulf
>



-- 
Greg Kurtzer
http://www.infiscale.com/
http://www.perceus.org/
http://www.caoslinux.org/


From ranjeet.rai at orkash.com  Fri Jun  5 02:28:30 2009
From: ranjeet.rai at orkash.com (ranjeet.rai at orkash.com)
Date: Fri, 05 Jun 2009 14:58:30 +0530
Subject: [Warewulf] Upgrading Perceus to 1.5.2
Message-ID: <4A28E53E.60905@orkash.com>

Hi,

I have configured CAOS NSA with Perceus 1.5.0 for my testing 2 node 
cluster. I am trying to upgrading perceus with it's latest available 
version i.e 1.5.2. I downloaded the tar.gz package  and configured and 
installed (./configure, make & make install).

Now, when I run the command #/etc/init.d/perceus restart/start/stop etc. 
I get the message "you need to initialize Perceus before starting the 
services !"

please help me to troubleshoot this message and run my perceus again.

Thanks in advance.

Best regards,

Ranjeet Rai


From griznog at gmail.com  Fri Jun  5 03:56:41 2009
From: griznog at gmail.com (John Hanks)
Date: Fri, 5 Jun 2009 06:56:41 -0400
Subject: [Warewulf] Upgrading Perceus to 1.5.2
In-Reply-To: <4A28E53E.60905@orkash.com>
References: <4A28E53E.60905@orkash.com>
Message-ID: <d98be35b0906050356o1d8944b2v8deba511f24d9b98@mail.gmail.com>

On Fri, Jun 5, 2009 at 5:28 AM,
ranjeet.rai at orkash.com<ranjeet.rai at orkash.com> wrote:
> Hi,
>
> I have configured CAOS NSA with Perceus 1.5.0 for my testing 2 node
> cluster. I am trying to upgrading perceus with it's latest available
> version i.e 1.5.2. I downloaded the tar.gz package  and configured and
> installed (./configure, make & make install).
>
> Now, when I run the command #/etc/init.d/perceus restart/start/stop etc.
> I get the message "you need to initialize Perceus before starting the
> services !"
>

Hmm, this implies it wasn't just me being stupid. I went through this
a few days ago on CentOS 5 and had the same issue. My impression was
that the make install step somehow foobar'd the --prefix and put new
database (and possibly other) paths into play. Rather than
troubleshoot, I just started over as mine was a throwaway test system.
But this does imply to me that there is a problem with the "make
install" step in 1.5.2 and where it thinks things should go.

You might find your old databases in the old location, I'd try poking
around /usr/local/var/lib/perceus and /var/lib/perceus (or related
paths) and see if you have multiple database localtions. In my case I
thought I did that correctly and backed it up but it turned out those
were empty (thus my opting for a do-over.) I can't prove now if I
backed up the wrong thing or if "make install" overwrote my db files
so other than "me too" I'm about as little help as you can get.

jbh


From astevens at infiscale.com  Fri Jun  5 07:38:13 2009
From: astevens at infiscale.com (astevens at infiscale.com)
Date: Fri, 5 Jun 2009 14:38:13 +0000
Subject: [Warewulf] Upgrading Perceus to 1.5.2
Message-ID: <391821016-1244212717-cardhu_decombobulator_blackberry.rim.net-419156927-@bxe1251.bisx.prod.on.blackberry>

On Caos NSA, to update the system, all you have to do is...

smart update
smart upgrade

From your root prompt, or you can update the system via sidekick as well, no need to mess with the rpm's or do anything else :)
------Original Message------
From: ranjeet.rai at orkash.com
Sender: warewulf-bounces at caoslinux.org
To: warewulf at caoslinux.org
ReplyTo: The Warewulf Cluster Toolkit
Subject: [Warewulf] Upgrading Perceus to 1.5.2
Sent: Jun 5, 2009 2:28 AM

Hi,

I have configured CAOS NSA with Perceus 1.5.0 for my testing 2 node 
cluster. I am trying to upgrading perceus with it's latest available 
version i.e 1.5.2. I downloaded the tar.gz package  and configured and 
installed (./configure, make & make install).

Now, when I run the command #/etc/init.d/perceus restart/start/stop etc. 
I get the message "you need to initialize Perceus before starting the 
services !"

please help me to troubleshoot this message and run my perceus again.

Thanks in advance.

Best regards,

Ranjeet Rai
_______________________________________________
Warewulf mailing list
Warewulf at caoslinux.org
http://lists.caosity.org/mailman/listinfo/warewulf


Sent via BlackBerry from T-Mobile

From ranjeet.rai at orkash.com  Sun Jun  7 21:13:07 2009
From: ranjeet.rai at orkash.com (ranjeet.rai at orkash.com)
Date: Mon, 08 Jun 2009 09:43:07 +0530
Subject: [Warewulf] Warewulf Digest, Vol 54, Issue 3
In-Reply-To: <mailman.1.1244289601.26636.warewulf@caoslinux.org>
References: <mailman.1.1244289601.26636.warewulf@caoslinux.org>
Message-ID: <4A2C8FD3.4020305@orkash.com>

Hi,

I tried upgrading via "smart update/upgrade, it resolves the package 
perceus 1.5.1. I feel the latest 1.5.2 is not updated in the image list 
or there may be some other issues.

Please help me to achieve this.

Best regards,

Ranjeet Rai

warewulf-request at caoslinux.org wrote:
> Send Warewulf mailing list submissions to
> 	warewulf at caoslinux.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> 	http://lists.caosity.org/mailman/listinfo/warewulf
> or, via email, send a message with subject or body 'help' to
> 	warewulf-request at caoslinux.org
>
> You can reach the person managing the list at
> 	warewulf-owner at caoslinux.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Warewulf digest..."
>
>
> Today's Topics:
>
>    1. Re: Upgrading Perceus to 1.5.2 (astevens at infiscale.com)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Fri, 5 Jun 2009 14:38:13 +0000
> From: astevens at infiscale.com
> Subject: Re: [Warewulf] Upgrading Perceus to 1.5.2
> To: "The Warewulf Cluster Toolkit" <warewulf at caoslinux.org>
> Message-ID:
> 	<391821016-1244212717-cardhu_decombobulator_blackberry.rim.net-419156927- at bxe1251.bisx.prod.on.blackberry>
> 	
> Content-Type: text/plain
>
> On Caos NSA, to update the system, all you have to do is...
>
> smart update
> smart upgrade
>
> >From your root prompt, or you can update the system via sidekick as well, no need to mess with the rpm's or do anything else :)
> ------Original Message------
> From: ranjeet.rai at orkash.com
> Sender: warewulf-bounces at caoslinux.org
> To: warewulf at caoslinux.org
> ReplyTo: The Warewulf Cluster Toolkit
> Subject: [Warewulf] Upgrading Perceus to 1.5.2
> Sent: Jun 5, 2009 2:28 AM
>
> Hi,
>
> I have configured CAOS NSA with Perceus 1.5.0 for my testing 2 node 
> cluster. I am trying to upgrading perceus with it's latest available 
> version i.e 1.5.2. I downloaded the tar.gz package  and configured and 
> installed (./configure, make & make install).
>
> Now, when I run the command #/etc/init.d/perceus restart/start/stop etc. 
> I get the message "you need to initialize Perceus before starting the 
> services !"
>
> please help me to troubleshoot this message and run my perceus again.
>
> Thanks in advance.
>
> Best regards,
>
> Ranjeet Rai
> _______________________________________________
> Warewulf mailing list
> Warewulf at caoslinux.org
> http://lists.caosity.org/mailman/listinfo/warewulf
>
>
> Sent via BlackBerry from T-Mobile
>
> ------------------------------
>
> _______________________________________________
> Warewulf mailing list
> Warewulf at caoslinux.org
> http://lists.caosity.org/mailman/listinfo/warewulf
>
>
> End of Warewulf Digest, Vol 54, Issue 3
> ***************************************
>
>   



From ranjeet.rai at orkash.com  Sun Jun  7 23:16:34 2009
From: ranjeet.rai at orkash.com (ranjeet.rai at orkash.com)
Date: Mon, 08 Jun 2009 11:46:34 +0530
Subject: [Warewulf] Testing application for HPC
In-Reply-To: <mailman.1.1244289601.26636.warewulf@caoslinux.org>
References: <mailman.1.1244289601.26636.warewulf@caoslinux.org>
Message-ID: <4A2CACC2.6030601@orkash.com>

Hi,

I have created a three node cluster with CAOS NSA and perceus 1.5.1. I 
want to test this cluster by running some other application on these 
parallel node. Please, help me to test this HPC.

I want to run this test to visualize the working of our cluster 
environment with any freely/openly available application.

Any support and guidance in this aspects will highly be appreciated.

Best regards,

Ranjeet Rai
Orkash Services Pvt Ltd

Mob: +91 9810499844

Tel: +91 124 2345773

www.orkash.com
... ensuring Assurance in uncertainty and complexity

This message including the attachments, if any, is a confidential 
business communication. If you are not the intended recipient it may be 
unlawful for you to read, copy, distribute, disclose or otherwise use 
the information in this e-mail. If you have received it in error or are 
not the intended recipient, please destroy it and notify the sender 
immediately. Thank you


From ranjeet.rai at orkash.com  Tue Jun  9 23:15:49 2009
From: ranjeet.rai at orkash.com (ranjeet.rai at orkash.com)
Date: Wed, 10 Jun 2009 11:45:49 +0530
Subject: [Warewulf] HPC testing with JAVA application
In-Reply-To: <mailman.1.1244462402.23237.warewulf@caoslinux.org>
References: <mailman.1.1244462402.23237.warewulf@caoslinux.org>
Message-ID: <4A2F4F95.3010207@orkash.com>

Hi,

I have created a 2 node cluster using perceus under Caos NSA. I want to 
run an application which is java based in combination with JPVM package.

This environment also requires jdk environment on cluster nodes. How can 
I configure jdk for cluster nodes?

Also, guidelines on testing the performance of this cluster will be 
appreciated.

Best regards,

Ranjeet Rai

Orkash Services Pvt Ltd

Mob: +91 9810499844

Tel: +91 124 2345773

www.orkash.com
... ensuring Assurance in uncertainty and complexity

This message including the attachments, if any, is a confidential 
business communication. If you are not the intended recipient it may be 
unlawful for you to read, copy, distribute, disclose or otherwise use 
the information in this e-mail. If you have received it in error or are 
not the intended recipient, please destroy it and notify the sender 
immediately. Thank you



From jsquyres at cisco.com  Thu Jun 11 19:06:44 2009
From: jsquyres at cisco.com (Jeff Squyres)
Date: Thu, 11 Jun 2009 19:06:44 -0700
Subject: [Warewulf] Perceus 1.5.2 build problem on RHEL4u4
Message-ID: <7D1819AA-9FD1-4BE5-B728-6FE347214B86@cisco.com>

I was trying to upgrade from Perceus 1.3.4 (gasp!) to 1.5.2 today.  My  
cluster head node is RHEL4u4 (unfortunately, can't change it :-( ).   
The perceus kernel failed to build -- here's a snipit of the failure  
(full config.log, config.out, and make.out attached):

...
   LD      arch/x86/lib/built-in.o
   AR      arch/x86/lib/lib.a
   LD      vmlinux.o
   MODPOST vmlinux.o
   GEN     .version
   CHK     include/linux/compile.h
   UPD     include/linux/compile.h
   CC      init/version.o
   LD      init/built-in.o
   LD      vmlinux
mm/built-in.o(.text+0x6f16): In function `test_clear_page_writeback':
: undefined reference to `____ilog2_NaN'
make[2]: *** [vmlinux] Error 1
make[2]: Leaving directory `/home/jsquyres/perceus/ 
perceus-1.5.2/3rd_party/_work/kernel/linux-2.6.28'
...
make[2]: Leaving directory `/home/jsquyres/perceus/ 
perceus-1.5.2/3rd_party/_work/kernel/linux-2.6.28'
cp _work/kernel/linux-2.6.28/arch/x86_64/boot/bzImage ../
cp: cannot stat `_work/kernel/linux-2.6.28/arch/x86_64/boot/bzImage':  
No such file or directory
make[1]: *** [kernel] Error 1
make[1]: Leaving directory `/home/jsquyres/perceus/ 
perceus-1.5.2/3rd_party'
make: *** [all-recursive] Error 1

Any ideas?  Is Perceus 1.5.2 supported on RHEL4u4?

Thanks!

-- 
Jeff Squyres
Cisco Systems
-------------- next part --------------
A non-text attachment was scrubbed...
Name: perceus-1.5.2.tar.bz2
Type: application/x-bzip2
Size: 26197 bytes
Desc: not available
Url : http://altruistic.infiscale.org/pipermail/perceus/attachments/20090611/62aec7c4/attachment.bz2 

From gmkurtzer at gmail.com  Thu Jun 11 20:11:03 2009
From: gmkurtzer at gmail.com (Greg Kurtzer)
Date: Thu, 11 Jun 2009 20:11:03 -0700
Subject: [Warewulf] Perceus 1.5.2 build problem on RHEL4u4
In-Reply-To: <7D1819AA-9FD1-4BE5-B728-6FE347214B86@cisco.com>
References: <7D1819AA-9FD1-4BE5-B728-6FE347214B86@cisco.com>
Message-ID: <571f1a060906112011w49d8101h371a956a12e44df8@mail.gmail.com>

Heya Jeff!

We have seen this before and IIRC, the version of binutils is too old
on RHEL4 to build recent kernels.

There are several kludges that I can think of, none of which I would
recommend on a public forum. ;)

Feel free to contact me off list and I can work with you directly on this.

Greg

On Thu, Jun 11, 2009 at 7:06 PM, Jeff Squyres<jsquyres at cisco.com> wrote:
> I was trying to upgrade from Perceus 1.3.4 (gasp!) to 1.5.2 today. ?My
> cluster head node is RHEL4u4 (unfortunately, can't change it :-( ). ?The
> perceus kernel failed to build -- here's a snipit of the failure (full
> config.log, config.out, and make.out attached):
>
> ...
> ?LD ? ? ?arch/x86/lib/built-in.o
> ?AR ? ? ?arch/x86/lib/lib.a
> ?LD ? ? ?vmlinux.o
> ?MODPOST vmlinux.o
> ?GEN ? ? .version
> ?CHK ? ? include/linux/compile.h
> ?UPD ? ? include/linux/compile.h
> ?CC ? ? ?init/version.o
> ?LD ? ? ?init/built-in.o
> ?LD ? ? ?vmlinux
> mm/built-in.o(.text+0x6f16): In function `test_clear_page_writeback':
> : undefined reference to `____ilog2_NaN'
> make[2]: *** [vmlinux] Error 1
> make[2]: Leaving directory
> `/home/jsquyres/perceus/perceus-1.5.2/3rd_party/_work/kernel/linux-2.6.28'
> ...
> make[2]: Leaving directory
> `/home/jsquyres/perceus/perceus-1.5.2/3rd_party/_work/kernel/linux-2.6.28'
> cp _work/kernel/linux-2.6.28/arch/x86_64/boot/bzImage ../
> cp: cannot stat `_work/kernel/linux-2.6.28/arch/x86_64/boot/bzImage': No
> such file or directory
> make[1]: *** [kernel] Error 1
> make[1]: Leaving directory `/home/jsquyres/perceus/perceus-1.5.2/3rd_party'
> make: *** [all-recursive] Error 1
>
> Any ideas? ?Is Perceus 1.5.2 supported on RHEL4u4?
>
> Thanks!
>
> --
> Jeff Squyres
> Cisco Systems
>
> _______________________________________________
> Warewulf mailing list
> Warewulf at caoslinux.org
> http://lists.caosity.org/mailman/listinfo/warewulf
>
>



-- 
Greg Kurtzer
http://www.infiscale.com/
http://www.perceus.org/
http://www.caoslinux.org/


From bernard at vanhpc.org  Mon Jun 15 17:59:44 2009
From: bernard at vanhpc.org (Bernard Li)
Date: Mon, 15 Jun 2009 17:59:44 -0700
Subject: [Warewulf] perceus module deactivate bug
Message-ID: <d4c731da0906151759t69ea79d2k69faa19838a233d9@mail.gmail.com>

Hi all:

Using Perceus 1.5.2.  Activating and subsequently deactivating the
same module on a node gives me this:

# perceus module activate provision init/node/node01
Perceus Module 'provision' has been enabled in 'init/node/node01'
# perceus module deactivate provision init/node/node01
WARNING: Inappropriate role name: init/node/node01
Perceus Module 'provision' has been disabled in 'init/node/node01'

The warning seems like a bug to me, patch as follows:

---cut---
Index: scripts/lib/Perceus/Interface/Cmdline.pm
===================================================================
--- scripts/lib/Perceus/Interface/Cmdline.pm	(revision 2114)
+++ scripts/lib/Perceus/Interface/Cmdline.pm	(working copy)
@@ -1587,8 +1587,8 @@
     }

     foreach my $r ( @roles_req ) {
-        if ( $r =~ /^([a-zA-Z0-9]+\/[a-zA-Z0-9]+)$/ ) {
-            &dprint("Adding '$r' to untainted roles to activate");
+        if ( $r =~ /^([a-zA-Z0-9]+(\/[a-zA-Z0-9]+)+)$/ ) {
+            &dprint("Adding '$r' to untainted roles to deactivate");
             push(@roles, $1);
         } else {
             &wprint("Inappropriate role name: $r");
---cut---

I also fixed a typo: "activate" --> "deactivate".

Pretty simple patch, I could attach it if needed.

Thanks,

Bernard


From gmkurtzer at gmail.com  Fri Jun 19 17:40:57 2009
From: gmkurtzer at gmail.com (Greg Kurtzer)
Date: Fri, 19 Jun 2009 17:40:57 -0700
Subject: [Warewulf] perceus module deactivate bug
In-Reply-To: <d4c731da0906151759t69ea79d2k69faa19838a233d9@mail.gmail.com>
References: <d4c731da0906151759t69ea79d2k69faa19838a233d9@mail.gmail.com>
Message-ID: <571f1a060906191740x4dc6dfbci4c91a15c365d9297@mail.gmail.com>

Got it. Thanks!

On Mon, Jun 15, 2009 at 5:59 PM, Bernard Li<bernard at vanhpc.org> wrote:
> Hi all:
>
> Using Perceus 1.5.2. ?Activating and subsequently deactivating the
> same module on a node gives me this:
>
> # perceus module activate provision init/node/node01
> Perceus Module 'provision' has been enabled in 'init/node/node01'
> # perceus module deactivate provision init/node/node01
> WARNING: Inappropriate role name: init/node/node01
> Perceus Module 'provision' has been disabled in 'init/node/node01'
>
> The warning seems like a bug to me, patch as follows:
>
> ---cut---
> Index: scripts/lib/Perceus/Interface/Cmdline.pm
> ===================================================================
> --- scripts/lib/Perceus/Interface/Cmdline.pm ? ?(revision 2114)
> +++ scripts/lib/Perceus/Interface/Cmdline.pm ? ?(working copy)
> @@ -1587,8 +1587,8 @@
> ? ? }
>
> ? ? foreach my $r ( @roles_req ) {
> - ? ? ? ?if ( $r =~ /^([a-zA-Z0-9]+\/[a-zA-Z0-9]+)$/ ) {
> - ? ? ? ? ? ?&dprint("Adding '$r' to untainted roles to activate");
> + ? ? ? ?if ( $r =~ /^([a-zA-Z0-9]+(\/[a-zA-Z0-9]+)+)$/ ) {
> + ? ? ? ? ? ?&dprint("Adding '$r' to untainted roles to deactivate");
> ? ? ? ? ? ? push(@roles, $1);
> ? ? ? ? } else {
> ? ? ? ? ? ? &wprint("Inappropriate role name: $r");
> ---cut---
>
> I also fixed a typo: "activate" --> "deactivate".
>
> Pretty simple patch, I could attach it if needed.
>
> Thanks,
>
> Bernard
> _______________________________________________
> Warewulf mailing list
> Warewulf at caoslinux.org
> http://lists.caosity.org/mailman/listinfo/warewulf
>



-- 
Greg Kurtzer
http://www.infiscale.com/
http://www.perceus.org/
http://www.caoslinux.org/


From agshew at gmail.com  Mon Jun 22 09:01:28 2009
From: agshew at gmail.com (Andrew Shewmaker)
Date: Mon, 22 Jun 2009 10:01:28 -0600
Subject: [Warewulf] debugging provisiond
In-Reply-To: <571f1a060906011342h20988a0bo27e9cbef6664c04b@mail.gmail.com>
References: <f4050c9e0906011039k30776809g48bab48507039782@mail.gmail.com>
	<f4050c9e0906011253l6d4d7631r52ca40e2b2d3099a@mail.gmail.com>
	<571f1a060906011342h20988a0bo27e9cbef6664c04b@mail.gmail.com>
Message-ID: <f4050c9e0906220901l54431b2ds880fad75f5ff95f7@mail.gmail.com>

On Mon, Jun 1, 2009 at 2:42 PM, Greg Kurtzer<gmkurtzer at gmail.com> wrote:
> On Mon, Jun 1, 2009 at 12:53 PM, Andrew Shewmaker <agshew at gmail.com> wrote:
> > On Mon, Jun 1, 2009 at 11:39 AM, Andrew Shewmaker <agshew at gmail.com> wrote:
> >> I'm having an issue with the passwdfile module where it isn't putting
> >> the contents of /etc/perceus/modules/passwdfile/all into /etc/passwd
> > ...
> >> What should I be looking at?  I must be doing something dumb.
> >
> > Like having perceus-provisiond installed in the vnfs, but with its
> > files listed in the hybridize file.  I didn't get symlinks for
> > provisiond, but I did for other stuff.
>
> Hello Andrew,
>
> Good catch... Now I wonder why!?
>
> If you notice anything further please let us know. Otherwise, please
> send the hybridized entries for provisiond so we can try and
> replicate.

Thanks for your response ... I didn't see it for a while because my
email filter hid it from view.  The good news is I finally found my
problem.  It had to do with a custom version of the umount script
that we are using with our perceus/cfengine integration.

I'm not sure why, but

HYBRIDIZE=`grep -v "^#" $HYBRIDIZE_FILE | sed -e 's/^\///g'`

was replaced with

HYBRIDIZE=`grep -v "^#" $HYBRIDIZE_FILE`

and that isn't good because then the find used for symlinking
is looking at the master node file system's root
instead of the vnfs.  Most of the files existed in both
places, but we have the occasional package which isn't.

-- 
Andrew Shewmaker


From chaz.whittemore at gmail.com  Wed Jun 24 09:23:40 2009
From: chaz.whittemore at gmail.com (Chaz Whittemore)
Date: Wed, 24 Jun 2009 12:23:40 -0400
Subject: [Warewulf] First cluster - some general caos/perceus questions
Message-ID: <86079c030906240923g5fdb0417t52fa4dee974e506@mail.gmail.com>

Hello all! I am setting up my first (production) cluster, and I had some
general questions I was hoping you all could help me with. II am using
Perceus/Warewulf/openmpi/slurm packages contained in caos NSA by default.

1) When running 1.0.8 on the master and all nodes, everything worked great.
However, after updating all of the packages via smart (including kernel,
slurm, and openmpi), some strange things happened. While slurm still works,
I now have to manually restart the slurmd on the each node after restarts to
get sinfo to see them as active. Also, mpirun seems to be broken now...
maybe this is due to a version mismatch b/c the vnfs didnt get updated?

2) Is there a way to run smart updates on the vnfs capsule? I am used to yum
and rpm with their root= and install-root parameters - is there an
equivalent for smart?

3) At the fancy new graphical Perceus splash - how do I reduce the wait time
from 30 seconds to something shorter?

4) I have configured the nodes to each have a scratch drive. I have
partritioned the drives into two partitions, a swap and an ext3 for /scratch
(formatted too) - and I uncommented the matching lines in the vnfs'
/etc/fstab. However, when the node starts up, I see the message "Special
device /dev/sda2 does not exist." - and upon further inspection, neither
partitin gets mounted. I CAN, however, mount them immediately after boot
with "mount /dev/sda2 /scratch" and "swapon /dev/sda1" and they work fine.
Is there something I am missing?

5) This is more of a netrworking question - maybe someone here has some
input.  I initially set up the small cluster on an umanaged soho 8 port
gigabit switch. But, after convincing the higher ups that we need a managed
switch - I installed it, only to find that the nodes are taking twice as
long to boot. Over three minuts each! (Only about 1:45 on the unmanaged
switch). Is there something I should be doing with my managed switch out of
the box to get it of comparible performance?

Thanks everyone - especially for your patience!!!
-chaz
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://altruistic.infiscale.org/pipermail/perceus/attachments/20090624/c7b3e3bb/attachment.html 

From gmkurtzer at gmail.com  Wed Jun 24 10:15:28 2009
From: gmkurtzer at gmail.com (Greg Kurtzer)
Date: Wed, 24 Jun 2009 10:15:28 -0700
Subject: [Warewulf] First cluster - some general caos/perceus questions
In-Reply-To: <86079c030906240923g5fdb0417t52fa4dee974e506@mail.gmail.com>
References: <86079c030906240923g5fdb0417t52fa4dee974e506@mail.gmail.com>
Message-ID: <571f1a060906241015x708507adj77ffd211425ed5ea@mail.gmail.com>

On Wed, Jun 24, 2009 at 9:23 AM, Chaz
Whittemore<chaz.whittemore at gmail.com> wrote:
> Hello all! I am setting up my first (production) cluster, and I had some
> general questions I was hoping you all could help me with. II am using
> Perceus/Warewulf/openmpi/slurm packages contained in caos NSA by default.

Hello, and welcome aboard! ;)

>
> 1) When running 1.0.8 on the master and all nodes, everything worked great.
> However, after updating all of the packages via smart (including kernel,
> slurm, and openmpi), some strange things happened. While slurm still works,
> I now have to manually restart the slurmd on the each node after restarts to
> get sinfo to see them as active. Also, mpirun seems to be broken now...
> maybe this is due to a version mismatch b/c the vnfs didnt get updated?

Yes, I think your right about the VNFS not being at the same version
as the master. Try this:

# perceus vnfs mount [vnfs name]
# smart -o rpm-root=/mnt/[vnfs name] upgrade
# perceus vnfs umount [vnfs name]

Please note that if the kernel version changes, you will have to
update the VNFS configuration in /etc/perceus/vnfs/[vnfs name]/config.

>
> 2) Is there a way to run smart updates on the vnfs capsule? I am used to yum
> and rpm with their root= and install-root parameters - is there an
> equivalent for smart?

Yep. See #1.

>
> 3) At the fancy new graphical Perceus splash - how do I reduce the wait time
> from 30 seconds to something shorter?

Edit the file /var/lib/perceus/tftp/pxelinux.cfg/default and change
300 to something smaller.

>
> 4) I have configured the nodes to each have a scratch drive. I have
> partritioned the drives into two partitions, a swap and an ext3 for /scratch
> (formatted too) - and I uncommented the matching lines in the vnfs'
> /etc/fstab. However, when the node starts up, I see the message "Special
> device /dev/sda2 does not exist." - and upon further inspection, neither
> partitin gets mounted. I CAN, however, mount them immediately after boot
> with "mount /dev/sda2 /scratch" and "swapon /dev/sda1" and they work fine.
> Is there something I am missing?

My guess is that the device drivers are taking some time to initialize
and are not blocking. There are several solutions. The easiest is to
kludge it with a "mount -a" in the /etc/rc.local.

>
> 5) This is more of a netrworking question - maybe someone here has some
> input.? I initially set up the small cluster on an umanaged soho 8 port
> gigabit switch. But, after convincing the higher ups that we need a managed
> switch - I installed it, only to find that the nodes are taking twice as
> long to boot. Over three minuts each! (Only about 1:45 on the unmanaged
> switch). Is there something I should be doing with my managed switch out of
> the box to get it of comparible performance?

Can you identify the point of the boot that is it causing the
slowness? During the provisioning itself, or timing out and retrying?

At first guess check to make sure the switch is setting duplexing and
auto-negotiating properly.

>
> Thanks everyone - especially for your patience!!!

Best of luck!


-- 
Greg Kurtzer
http://www.infiscale.com/
http://www.perceus.org/
http://www.caoslinux.org/


From gwleong at gmail.com  Wed Jun 24 15:51:20 2009
From: gwleong at gmail.com (Gary Leong)
Date: Wed, 24 Jun 2009 15:51:20 -0700
Subject: [Warewulf] pmod
Message-ID: <ae974e900906241551q7ed7aee3x7d12bd90ada10d5d@mail.gmail.com>

Does anybody know the format for pmod?  I can insert the module manually,
but just wanted to know if it was better to use the actual import
functionality on perceus.

Gary
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://altruistic.infiscale.org/pipermail/perceus/attachments/20090624/de81c93b/attachment.html 

From WSChoong at lbl.gov  Wed Jun 24 23:43:28 2009
From: WSChoong at lbl.gov (Woon-Seng Choong)
Date: Wed, 24 Jun 2009 23:43:28 -0700
Subject: [Warewulf] Problem registering perceus
Message-ID: <f10f8f2444a4.4a42ba20@lbl.gov>

I am trying to install Caos NSA and perceus. Let me just say that I have no problem registering with perceus, but I keep getting error registering perceus. I am trying to setup a cluster for a collaborator in Malaysia, and initially I don't even connect to the outside world. Every time I try to run a perceus command, it jumps into the registration. And when I try to boot up a node, it doesn't come up completely and I don't know if this is due to not registering. Is there a way to bypass the registration and make the node boot up properly.

     Thank you
     Seng




From stefan at mdy.univie.ac.at  Thu Jun 25 00:42:13 2009
From: stefan at mdy.univie.ac.at (Stefan Boresch)
Date: Thu, 25 Jun 2009 09:42:13 +0200
Subject: [Warewulf] First cluster - some general caos/perceus questions
In-Reply-To: <86079c030906240923g5fdb0417t52fa4dee974e506@mail.gmail.com>
References: <86079c030906240923g5fdb0417t52fa4dee974e506@mail.gmail.com>
Message-ID: <20090625074213.GF19565@loop.mdy.univie.ac.at>

Hi Chaz,

Greg already gave you the authoritative answers ...

On Wed, Jun 24, 2009 at 12:23:40PM -0400, Chaz Whittemore wrote:
> 
> 4) I have configured the nodes to each have a scratch drive. I have
> partritioned the drives into two partitions, a swap and an ext3 for /scratch
> (formatted too) - and I uncommented the matching lines in the vnfs'
> /etc/fstab. However, when the node starts up, I see the message "Special
> device /dev/sda2 does not exist." - and upon further inspection, neither
> partitin gets mounted. I CAN, however, mount them immediately after boot
> with "mount /dev/sda2 /scratch" and "swapon /dev/sda1" and they work fine.
> Is there something I am missing?
> 

I just want to confirm Greg's suspicion here as a fact. To speed up
the boot process on the nodes, the are booting in 'fast' mode, not
waiting for the dynamic devices to  have "settled", hence your mount -a
comes too early.  If you are in a hurry, definitely follow Greg's
advice and (mis)use rc.local.  Otherwise look at the start up scripts
to locate the trouble area.  I had this 'fixed' in a local version
of mine before I lost the changes in a complete reinstall :-(  Since
then I am happy with the rc.local kludge ...

Good luck,

Stefan

-- 
Stefan Boresch
Institute for Computational Biological Chemistry
University of Vienna, Waehringerstr. 17       A-1090 Vienna, Austria
Phone: -43-1-427752715                        Fax:   -43-1-427752790