[Linux-HA] question: external/ssh stonith to poweroff badnode via xen-host

Serge Dubrouski sergeyfd at gmail.com
Mon Mar 3 09:57:15 MST 2008 

Configuration looks right to me, I even tested it and it worked fine
on my test cluster. So hints are obvious:

1. Check that you really put that script on a second node and made it
executable.
2. Nodes should be able to ping each other. That programmed in a
"status" function.



On Mon, Mar 3, 2008 at 9:16 AM, Lino Moragon <lino.moragon at highspeed.li> wrote:
> Hi,
>
>  I'm using now the most actual xen0 stonith plugin, that Serge attachted
>  the 2008-02-28 to this thread.
>  I thought I configured everything correct but it seems that the stonith
>  clone cannot be started on my 2nd node.
>  I must admit I configured the Clone via hb_gui but I still have some issues.
>  As for these reasons the stonith plugin only works to reset my 2nd node
>  and not vice versa.
>  The version I am using is: heartbeat-2.1.2-3 of the Centos5.1 repository.
>
>  Has anyone a clue where the failure could be?
>
>  To this email I attach the cib.xml
>  Below follows the cib stonith section, crm_verfy and error msgs.
>
>  Perhaps Serge, you could give me a hint? What did you do differently
>  than me?
>
>  I would be very glad for any hints and clues.
>
>  Thanks in advance,
>   Lino
>
>  node1: mysql1
>  node2: mysql2
>  xen-host: simulator
>
>        <clone id="DoFencing">
>          <instance_attributes id="DoFencing_instance_attrs">
>            <attributes>
>              <nvpair id="DoFencing_clone_max" name="clone_max" value="2"/>
>              <nvpair id="DoFencing_clone_node_max" name="clone_node_max"
>  value="1"/>
>              <nvpair id="DoFencing_target_role" name="target_role"
>  value="started"/>
>            </attributes>
>          </instance_attributes>
>          <primitive class="stonith" type="external/xen0"
>  provider="heartbeat" id="child_DoFencing">
>            <instance_attributes id="child_DoFencing_instance_attrs">
>              <attributes>
>                <nvpair name="target_role"
>  id="child_DoFencing_target_role" value="started"/>
>                <nvpair name="hostlist"
>  id="5525c381-5956-4564-af3d-2bc7b547812a" value="mysql1:mysql1.cfg
>  mysql2:mysql2.cfg"/>
>                <nvpair id="65feeaf5-501f-4648-a155-83b79b587fbf"
>  name="dom0" value="simulator"/>
>              </attributes>
>            </instance_attributes>
>          </primitive>
>        </clone>
>
>  If I use crm_verify I get the following results:
>  =============================================================
>  crm_verify[6169]: 2008/03/03_17:06:55 WARN: unpack_rsc_op: Processing
>  failed op (child_DoFencing:0_start_0) on mysql2
>  crm_verify[6169]: 2008/03/03_17:06:55 WARN: unpack_rsc_op: Handling
>  failed start for child_DoFencing:0 on mysql2
>  crm_verify[6169]: 2008/03/03_17:06:55 WARN: unpack_rsc_op: Processing
>  failed op (child_DoFencing:1_start_0) on mysql2
>  crm_verify[6169]: 2008/03/03_17:06:55 WARN: unpack_rsc_op: Handling
>  failed start for child_DoFencing:1 on mysql2
>  =============================================================
>
>  Furthermore I get the following Errors in my log:
>  =============================================================
>  Mar  3 16:29:42 mysql2 crmd: [1478]: ERROR: process_lrm_event: LRM
>  operation child_DoFencing:0_start_0 (call=22, rc=1) Error unknown error
>  Mar  3 16:29:46 mysql2 crmd: [1478]: ERROR: process_lrm_event: LRM
>  operation child_DoFencing:1_start_0 (call=24, rc=1) Error unknown error
>  Mar  3 16:35:33 mysql2 crmd: [1478]: ERROR: process_lrm_event: LRM
>  operation child_DoFencing:1_start_0 (call=28, rc=1) Error unknown error
>  Mar  3 16:46:40 mysql2 crmd: [1477]: ERROR: process_lrm_event: LRM
>  operation child_DoFencing:0_start_0 (call=10, rc=1) Error unknown error
>  Mar  3 16:46:45 mysql2 crmd: [1477]: ERROR: process_lrm_event: LRM
>  operation child_DoFencing:1_start_0 (call=12, rc=1) Error unknown error
>  ==============================================================
>
>
>
>
>
>  Serge Dubrouski schrieb:
>  > Attached.
>  >
>  > On Thu, Feb 28, 2008 at 3:35 AM, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
>  >> Hi Serge,
>  >>
>  >>
>  >>  On Tue, Feb 26, 2008 at 09:46:14AM -0700, Serge Dubrouski wrote:
>  >>  > Dejan -
>  >>  >
>  >>  > I found a compromise :-) Attached is a version of that plugin that
>  >>  > supports following parameters:
>  >>  >
>  >>  > 1. hostlist. A string that has a list of controlled nodes separated by
>  >>  > space or commas. A required parameter. In a simple form its just a
>  >>  > list of nodes. If one needs to use non-standard Xen configuration he
>  >>  > can use extended form of this parameter: "node1_name:config1_file
>  >>  > node2_name:config2_file". If config file isn't given it defaults to
>  >>  > /etc/xen/node_name.cfg
>  >>  >
>  >>  > 2. Dom0. Name of Dom0 Xen node. A required parameter.
>  >>  >
>  >>  > 3. ssh_command. SSH command that is used to ssh from DomU to Dom0.
>  >>  > Defaults to "/usr/bin/ssh -q -x -n -l root". If one wants to use SSH
>  >>  > keys for higher security he needs to use this parameter.
>  >>
>  >>  This is not necessary. One can setup everything needed in
>  >>  ~/.ssh/config on a per host basis, i.e. key or user to connect
>  >>  with. Since the plugin always runs as root, you might leave out
>  >>  '-l root' as well.
>  >>
>  >>
>  >>  > So in this form this plugin can be configured as a clone or as a set
>  >>  > of resources and location constraints.
>  >>  >
>  >>  >
>  >>  > I'd be very pleased if this plugin gets its way into Linux-HA distribution.
>  >>  >
>  >>
>  >>  Sure. Could you please just drop the ssh_command parameter.
>  >>
>  >>  Many thanks for the contribution.
>  >>
>  >>  Cheers,
>  >>
>  >>  Dejan
>  >>
>  >>
>  >>
>  >>  >
>  >>  > On Tue, Feb 26, 2008 at 8:45 AM, Serge Dubrouski <sergeyfd at gmail.com> wrote:
>  >>  > >
>  >>  > > On Mon, Feb 25, 2008 at 4:02 PM, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
>  >>  > >  > Hi,
>  >>  > >  >
>  >>  > >  >
>  >>  > >  >  On Mon, Feb 25, 2008 at 12:17:40PM -0700, Serge Dubrouski wrote:
>  >>  > >  >  > On Mon, Feb 25, 2008 at 12:10 PM, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
>  >>  > >  >  > > Hi,
>  >>  > >  >  > >
>  >>  > >  >  > >
>  >>  > >  >  > >  On Mon, Feb 25, 2008 at 11:27:38AM -0700, Serge Dubrouski wrote:
>  >>  > >  >  > >  > I would love to do that and already tried it. Though we didn't come to
>  >>  > >  >  > >  > agreement on how configuration parameters should look like.
>  >>  > >  >  > >
>  >>  > >  >  > >  Why? Was there a discussion on the list about it? The
>  >>  > >  >  > >  configuration is a bit unusual. Other stonith agents take named
>  >>  > >  >  > >  parameters. Though this kind of configuration also works, I'd
>  >>  > >  >  > >  prefer something similar to the others, e.g.
>  >>  > >  >  >
>  >>  > >  >  > Yes there was a discussion
>  >>  > >  >  > http://lists.community.tummy.com/pipermail/linux-ha-dev/2007-February/
>  >>  > >  >
>  >>  > >  >  It's a long one and peters out inconclusively.
>  >>  > >  >
>  >>  > >  >
>  >>  > >  >  > See "new stonith external plugin". The config parameter hostslist is
>  >>  > >  >  > actually derived from original ssh plugin. I needed to have a full
>  >>  > >  >  > list of all controlled nodes and preferred to have it as one
>  >>  > >  >  > parameter.
>  >>  > >  >  >
>  >>  > >  >  > >
>  >>  > >  >  > >  hostname dom0 (or xenhost) config
>  >>  > >  >  > >
>  >>  > >  >  >
>  >>  > >  >  > That would work if I needed just a dom0 host, but I also need a list
>  >>  > >  >  > of controlled nodes and probably configuration files.
>  >>  > >  >
>  >>  > >  >  That's why you can have several instances of a stonith resource
>  >>  > >  >  (see e.g. external/ipmi). Each of them would run with different
>  >>  > >  >  parameters. What I meant was:
>  >>  > >  >
>  >>  > >  >  hostname: xen vm
>  >>  > >  >  dom0: xen dom0
>  >>  > >  >  config: vm configuration file
>  >>  > >  >
>  >>  > >
>  >>  > >  That's possible and easy to do but I'm not sure that it'll be better.
>  >>  > >  Current version allows to configure a clone. New version would require
>  >>  > >  configuring a separate resource for each node and creating location
>  >>  > >  constraints for each of them. Per my opinion that would be more
>  >>  > >  complex configuration. May be I'm missing something.
>  >>  > >
>  >>  >
>  >>  >
>  >>  >
>  >>  > --
>  >>  > Serge Dubrouski.
>  >>
>  >>
>  >>
>  >>> _______________________________________________
>  >>  > Linux-HA mailing list
>  >>  > Linux-HA at lists.linux-ha.org
>  >>  > http://lists.linux-ha.org/mailman/listinfo/linux-ha
>  >>  > See also: http://linux-ha.org/ReportingProblems
>  >>
>  >>  --
>  >>  Dejan
>  >>
>  >>
>  >> _______________________________________________
>  >>  Linux-HA mailing list
>  >>  Linux-HA at 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 at 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 at lists.linux-ha.org
>  http://lists.linux-ha.org/mailman/listinfo/linux-ha
>  See also: http://linux-ha.org/ReportingProblems
>



-- 
Serge Dubrouski.

More information about the Linux-HA mailing list