Hello Florian,
Thanks again for your response, as for the build setup we are using
the following:
GCC Version:4.4
export PREFIX=/usr
export LCRSODIR=$PREFIX/libexec/lcrso
export CLUSTER_USER=hacluster
export CLUSTER_GROUP=haclient
getent group ${CLUSTER_GROUP} >/dev/null || groupadd -r ${CLUSTER_GROUP}
getent passwd ${CLUSTER_USER} >/dev/null || useradd -r -g
${CLUSTER_GROUP} -d /var/lib/heartbeat/cores/hacluster -s
/sbin/nologin -c "cluster user" ${CLUSTER_USER}
* Reusable Cluster Components (glue): ./configure --prefix=$PREFIX
--with-daemon-user=${CLUSTER_USER}
--with-daemon-group=${CLUSTER_GROUP} (maybe
--enable-fatal-warnings=no)
* Resource Agents: ./configure --prefix=$PREFIX
* Corosync: ./configure --prefix=$PREFIX --with-lcrso-dir=$LCRSODIR
* OpenAIS: same as corosync
* pcmk: ./configure --prefix=$PREFIX --with-lcrso-dir=$LCRSODIR
--with-corosync --with-cman
* Cluster 3: ./configure --prefix=/usr --without_rgmanager
--corosyncincdir=/usr/include/corocync --corosynclibdir=/usr/lib
openaisincdir=/usr/include/openais --corosyncbin=/usr/sbin/corosync
>> you are effectively trying to duplicate the work of a
>> distribution
Then it should work, unless I erred in the installation somwhere.
>> that would easily explain all runtime errors you've run into
I only experienced two of them, and the first one was fixed with Andrew's patch.
>> I'm also a bit baffled by your apparent requirement to be able to put
>> this "on any *nix based system" -- rest assured, getting this stack to
>> run on Solaris is a much bigger challenge than it is on Linux.
We only have to work our way through the same problem once. Next time
we will know exactly how to get
around it...
>> Finally, even though you have asked here multiple times about
>> dual-Primary DRBD, OCFS2, DLM, CMAN and friends, the only explanation
>> for _why_ you want this setup that I've found on the list is that you
>> want it for shared configuration files.
You missed the email expalining that we are only testing the entire
setup for simple files
at the moment, but will be making use of XEN, live migration etc... As
you know, there is
a lot happening with the entire spectrum (Cluster3, pcmk etc..) in
terms of development,
and it makes more sense for us at the moment to adopt the unstable in
parallel with
the what is included in the distro.
Florian, you have absolutely no idea who it is you're talking to. I'm
the calmest individual
it's almost eerie. How can you possible know what it takes to offend me?
Please sir, I am just trying to move forward with the project.
Kind Regards,
Nick.
On Mon, Nov 7, 2011 at 6:29 PM, Florian Haas <florian.h...@hastexo.com> wrote:
> On 2011-11-07 23:49, Nick Khamis wrote:
>> Hello Florian,
>>
>> Thank you so much for your response.
>>
>>>> I must confess I've lost track (and count) of all the threads you've
>>>> opened here on this list and others.
>>
>> I'm not sure if you're referring to this run-time error or previous
>> ones, and frankly I am not
>> sure how this is relevant.
>>
>> We are only using Debian for the prototype, and could be moving to any
>> *nix based system, and for this reason we are building everything from
>> source.
>> We should in theory be able to do this.... So no backport or distro patched
>> version would work.
>>
>>>> As for time pressure, I believe several people on this list have
>>>> informed you about various options to get professional assistance on
>>>> this, so I'll not rehash those here.
>>
>> Thanks. We're just looking to help ourselves and others compile and run,
>> pcmk+corosync using cman's standard dlm on an ocfs2 filesystem.
>> All built from source.
>>
>> Think about it, you're ridiculing us for helping in ironing out the bugs.
>
> Excuse me?
>
> OK, everyone has a right to be offended. So I'll not make any statement
> as to the alleged ridicule. That said, and I deliberately put it this
> bluntly, grow some thicker skin.
>
> In what you do (build everything from source, selecting versions of your
> own choosing, and in the case of Pacemaker even building straight out of
> git HEAD), you are effectively trying to duplicate the work of a
> distribution. Surely you're agreeing with me that the integrated effort
> required for such an undertaking is quite substantial. Therefore, I find
> it a bit surprising that you are doing this, when you could be using a
> stack provided by a specific distribution, or by a specific specialized
> vendor. As for your runtime errors -- I have no idea what QE/QA you run.
> I don't have a clue what GCC you're building with, I don't even know
> what configure options you set. For all I know, your build system could
> be perfect -- or it could be riddled with faults and problems that would
> easily explain all runtime errors you've run into, and then some. I just
> don't know. For distro packages, I'd know that your binary was built in
> the same environment as my binary. Thus, if you ran into an error with
> that package, I should be able to reproduce it. Reproducing problems in
> packages you rolled from source, not even released, with unknown
> configuration? I don't have a snowball's chance in hell.
>
> I'm also a bit baffled by your apparent requirement to be able to put
> this "on any *nix based system" -- rest assured, getting this stack to
> run on Solaris is a much bigger challenge than it is on Linux. In fact,
> I don't know if anyone regularly builds and tests Pacemaker on anything
> but Linux (except Andrew, who builds on Mac OS X, but I doubt he'd
> recommend that platform for production use).
>
> Finally, even though you have asked here multiple times about
> dual-Primary DRBD, OCFS2, DLM, CMAN and friends, the only explanation
> for _why_ you want this setup that I've found on the list is that you
> want it for shared configuration files. (Correct me if I'm wrong, but
> that's what you said in an email that was apparently inadvertently
> posted to this list, on Oct 25.) And if that's your sole reason for
> wanting a cluster file system, then sorry, that's a terrible idea. And
> Andreas even told you so. You could have solved this with NFS, or
> Csync2, or even rsync two weeks ago.
>
>> Please, now for an answer we could use.
>
> Hope the above helps. Then again, like I said, you already got an answer
> I humbly suggest you should have used. If you're asking for help, you
> get it, choose to ignore it, then scramble, and then complain when
> someone calls it out, again: you have the right to be offended.
>
> Florian
>
>
> _______________________________________________
> Linux-HA mailing list
> Linux-HA@lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
_______________________________________________
Linux-HA mailing list
Linux-HA@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
