Andreas,

> Pacemaker also sends the cluster node name to fence to the stonith
> agent, that is interpreted as "port". I am sure you read the description
> about how to use this agent and how it tries to find the correct EC2
> instance name to fence, using the node name.

Yes, I read and believe I understand what little documentation exists in
the source code (shell script) itself.  If there is additional
documentation somewhere, likely I've missed it somehow.  I'm not new to
shell scripting so I can see how the script obtains the actual EC2 instance
name given the "port" variable.  That's not a problem.  I do believe there
is an issue however with the script receiving the port/node information in
my configuration.

Here's a snip from /var/log/cluster/corosync.log:

Oct 08 20:34:06 [26569] node1 stonith-ng:    error: log_operation:
ec2-fencing: Required parameter 'INSTANCE' missing (-h for usage)

That "instance" parameter is calculated from "port", which seems to be
empty - not passed from Pacemaker.  I also modified the fence_ec2 script to
echo the port variable to a text file and confirmed that it is indeed empty.

I'm sure there are others using fence_ec2 so I attribute this issue to my
configuration somewhere... I just need to hone in on where I goofed the
configuration.  Everything seems to be working great, except for stonith.

-Kevin


On Mon, Oct 8, 2012 at 4:17 PM, Andreas Kurz <[email protected]> wrote:

> On 10/08/2012 10:44 PM, Kevin F. La Barre wrote:
> > Thank you Andreas, your input does further my understanding of how this
> is
> > supposed to work; however, I'm still unclear about your statement "where
> > the nodenames are sent as "port" parameter".  Specifically, how is the
> port
> > parameter passed from Pacemaker to the fencing script?  Should "port" be
> > specified in the primitive declaration, and if so, since it's dynamic we
> > cannot simply set it to a static node-name, right?  Also, if you look at
> > the code for fence_ec2 provided in the link within my original email the
> > port variable is essentially empty, port="".  If we leave this empty the
> > script will never receive the node name, but we cannot statically
> populate
> > the variable.  I would think that node A or node B would be the ones
> > fencing Node C.  Node C after-all may not be able to communicate.
>
> Pacemaker also sends the cluster node name to fence to the stonith
> agent, that is interpreted as "port". I am sure you read the description
> about how to use this agent and how it tries to find the correct EC2
> instance name to fence, using the node name.
>
> Regards,
> Andreas
>
>
> >
> > The above comes from the fact that the node/port is not being passed in
> my
> > configuration for some reason.  I'm seeing errors to the effect that
> > "INSTANCE" (the ec2 instance, aka port) is not being specified.
> >
> > Any help is much appreciated!
> >
> >
> > Respectfully,
> > Kevin
> >
> >
> > On Mon, Oct 8, 2012 at 2:10 PM, Andreas Kurz <[email protected]>
> wrote:
> >
> >> On 10/06/2012 08:45 AM, Kevin F. La Barre wrote:
> >>> I'm trying to get the fence_ec2 agent (link below) working and a bit
> >>> confused on how it should be configured.  I have modified the agent
> with
> >>> the EC2 key and cert, region, etc.  The part of confused about is the
> >>> "port" argument and how it's supposed to work.  Am I supposed to
> hardcode
> >>> the uname into the "port" variable or is this somehow passed into the
> >>> script as an argument?   If I hardcode it, I don't understand how
> >> Pacemaker
> >>> passes on the information as to which node to kill.  Versions and
> config.
> >>> details follow.
> >>>
> >>> I apologize if this has been vague.  Please let me know if you need
> more
> >>> information.
> >>>
> >>> Fencing agent:
> >>>
> >>
> https://github.com/beekhof/fence_ec2/blob/392a146b232fbf2bf2f75605b1e92baef4be4a01/fence_ec2
> >>>
> >>> crm configure primitive ec2-fencing stonith::fence_ec2 \
> >>> params action="reboot" \
> >>> op monitor interval="60s"
> >>
> >> try something like:
> >>
> >> primitive stonith_my-ec2-nodes stonith:fence_ec2 \
> >> params ec2-home="/root/.ec2" pcmk_host_check="static-list"
> >> pcmk_host_list="myec2-01 myec2-02" \
> >> op monitor interval="600s" timeout="300s" \
> >> op start start-delay="30s" interval="0"
> >>
> >>
> >> ... where the nodenames are sent as "port" paramter.
> >>
> >> Regards,
> >> Andreas
> >>
> >> --
> >> Need help with Pacemaker?
> >> http://www.hastexo.com/now
> >>
> >>>
> >>> Corosync v1.4.1
> >>> Pacemaker v1.1.7
> >>> CentOS 6.2
> >>>
> >>> -Kevin
> >>> _______________________________________________
> >>> Linux-HA mailing list
> >>> [email protected]
> >>> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> >>> See also: http://linux-ha.org/ReportingProblems
> >>>
> >>
> >>
> >>
> >>
> >>
> >> _______________________________________________
> >> Linux-HA mailing list
> >> [email protected]
> >> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> >> See also: http://linux-ha.org/ReportingProblems
> >>
> > _______________________________________________
> > Linux-HA mailing list
> > [email protected]
> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > See also: http://linux-ha.org/ReportingProblems
> >
>
>
> --
> Need help with Pacemaker?
> http://www.hastexo.com/now
>
>
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to