[Pacemaker] possible regex error in "pcs resource enable/disable"
Bob Haxo
bhaxo at sgi.com
Thu Feb 20 19:03:36 EST 2014
On Fri, 2014-02-21 at 10:46 +1100, Andrew Beekhof wrote:
> On 21 Feb 2014, at 10:25 am, Bob Haxo <bhaxo at sgi.com> wrote:
>
> > Andrew,
> >
> >> From my vantage point, it will be most unfortunate if pcs and crm cannot
> > both be used for routine start and stop of resources, and other routine
> > tasks.
>
> Agreed.
> AFAICS, it's crmsh thats creating the duplicate entries though.
> Not much pcs can do about that.
Yes, it is the crm that creates the duplicate entries. I have included
below the effect of pcs disable+enable and crm stop on the xml.
Sooo, seems that we need to kick this over the fence to the crmsh
folks ... or the SUSE folks, if they are maintaining crmsh.
------------------------------
mixing crm and pcs cmds
------------------------------
[root at mici-admin ~]# pcs resource disable libvirtd-clone
[root at mici-admin ~]# cibadmin --query --scope configuration > cib-config-pcs-disable-20140220-0855.xml
[root at mici-admin ~]# diff -u cib-config-initial-enabled-20140220-0854.xml cib-config-pcs-disable-20140220-0855.xml
--- cib-config-initial-enabled-20140220-0854.xml 2014-02-20 10:54:18.000000000 -0600
+++ cib-config-pcs-disable-20140220-0855.xml 2014-02-20 10:54:50.000000000 -0600
@@ -74,7 +74,9 @@
<nvpair id="libvirtd-interleave" name="interleave" value="true"/>
<nvpair id="libvirtd-ordered" name="ordered" value="true"/>
</meta_attributes>
- <meta_attributes id="libvirtd-clone-meta_attributes"/>
+ <meta_attributes id="libvirtd-clone-meta_attributes">
+ <nvpair id="libvirtd-clone-meta_attributes-target-role" name="target-role" value="Stopped"/>
+ </meta_attributes>
</clone>
<primitive class="ocf" id="virt" provider="heartbeat" type="VirtualDomain">
<instance_attributes id="virt-instance_attributes">
[root at mici-admin ~]# pcs resource enable libvirtd-clone
[root at mici-admin ~]# cibadmin --query --scope configuration > cib-config-pcs-enable-20140220-0900.xml
[root at mici-admin ~]# diff cib-config-initial-enabled-20140220-0854.xml cib-config-pcs-enable-20140220-0900.xml
[root at mici-admin ~]# <no diff>
[root at mici-admin ~]# crm resource stop libvirtd-clone
[root at mici-admin ~]# cibadmin --query --scope configuration > cib-config-crm-stop-20140220-0904.xml
[root at mici-admin ~]# diff -u cib-config-initial-enabled-20140220-0854.xml cib-config-crm-stop-20140220-0904.xml
--- cib-config-initial-enabled-20140220-0854.xml 2014-02-20 10:54:18.000000000 -0600
+++ cib-config-crm-stop-20140220-0904.xml 2014-02-20 11:04:34.000000000 -0600
@@ -73,8 +73,11 @@
<meta_attributes id="libvirtd-clone-meta">
<nvpair id="libvirtd-interleave" name="interleave" value="true"/>
<nvpair id="libvirtd-ordered" name="ordered" value="true"/>
+ <nvpair id="libvirtd-clone-meta-target-role" name="target-role" value="Stopped"/>
+ </meta_attributes>
+ <meta_attributes id="libvirtd-clone-meta_attributes">
+ <nvpair id="libvirtd-clone-meta_attributes-target-role" name="target-role" value="Stopped"/>
</meta_attributes>
- <meta_attributes id="libvirtd-clone-meta_attributes"/>
</clone>
<primitive class="ocf" id="virt" provider="heartbeat" type="VirtualDomain">
<instance_attributes id="virt-instance_attributes">
Bob Haxo
>
> >
> > I am using pcs (almost exclusively) for my RHEL6.5 Pacemaker port.
> >
> > But, I am **REALLY** hoping that I can present a relatively consistent
> > crmsh interface to the consumers of my HA work (customers, in-house
> > engineers) with RHEL and SLES installations. At least the in-house
> > engineers find the HA very confusing. This sort of a break will make it
> > impossible to ship with crmsh, thus continuing the confusion of two
> > distro specific interfaces.
> >
> > Regards,
> > Bob Haxo
> >
> >
> > On Fri, 2014-02-21 at 09:26 +1100, Andrew Beekhof wrote:
> >> On 21 Feb 2014, at 4:30 am, Bob Haxo <bhaxo at sgi.com> wrote:
> >>
> >>> Andrew, Lars,
> >>>
> >>> Yes, I have determined that this error is the result of mixing crmsh
> >> stop/start with pcs disable/enable (or maybe pcs stop/start mis-usage) commands.
> >>
> >> Specifically it will be when you use pcs first and crmsh afterwards.
> >> I just checked, pcs (which uses crm_resource) does the right thing if
> >> either one exists (ie. doesn't create a second entry), but necessarily
> >> bails out if both do.
> >>
> >> (libvirtd-clone-meta_attributes is the one created by crm_resource/pcs)
> >>
> >>>
> >>> I've started to respond with how this happens, but have been pulled
> >> off to a higher priority task. Back when I have the resolved.
> >>>
> >>> Bob Haxo
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> On Thu, 2014-02-20 at 07:56 -0800, Bob Haxo wrote:
> >>>> Andrew,
> >>>>
> >>>>
> >>>>
> >>>> I think that this is what you identified:
> >>>>
> >>>>
> >>>>
> >>>> This error appears to be related to the two instances of <meta_attributes> with
> >>>> libvirtd-clone, both which contain a name="target-role" value="Stopped".
> >>>> Deleted one of these entries and the error did not occur.
> >>>>
> >>>>
> >>>>
> >>>> <clone id="libvirtd-clone">
> >>>> <primitive class="lsb" id="libvirtd" type="libvirtd">
> >>>> <operations>
> >>>> <op id="libvirtd-monitor-interval-15" interval="15" name="monitor" start-delay="15" timeout="15"/>
> >>>> <op id="libvirtd-start-interval-0" interval="0" name="start" on-fail="restart" timeout="15"/>
> >>>> <op id="libvirtd-stop-interval-0" interval="0" name="stop" on-fail="ignore" timeout="15"/>
> >>>> </operations>
> >>>> <meta_attributes id="libvirtd-meta_attributes"/>
> >>>> </primitive>
> >>>> <meta_attributes id="libvirtd-clone-meta">
> >>>> <nvpair id="libvirtd-interleave" name="interleave" value="true"/>
> >>>> <nvpair id="libvirtd-ordered" name="ordered" value="true"/>
> >>>> <nvpair id="libvirtd-clone-meta-target-role" name="target-role" value="Stopped"/>
> >>>> </meta_attributes>
> >>>> <meta_attributes id="libvirtd-clone-meta_attributes">
> >>>> <nvpair id="libvirtd-clone-meta_attributes-target-role" name="target-role" value="Stopped"/>
> >>>> </meta_attributes>
> >>>> </clone>
> >>>>
> >>>>
> >>>>
> >>>> I do occasionally use a crmsh command, mostly the "crm resource cleanup <resource>" ...
> >>>> the pcs version of this seems to never cleanup the remote system entries. I classify
> >>>> this as a bug that I have not reported.
> >>>>
> >>>>
> >>>>
> >>>> So, yes, how the duplicate entry got there is probably the crux of the issue. And
> >>>> I have no answer. I have not used crmsh to create resources (the creates are
> >>>> scripted using pcs cmds). Perhaps I used a "crm stop" rather than a "pcs disable".
> >>>>
> >>>>
> >>>>
> >>>> Regardless, even if RH is not going to support crmsh, at least RH needs to be sensitive
> >>>> to the use of crmsh with their PCS Pacemaker implementation.
> >>>>
> >>>>
> >>>>
> >>>> I'll test/investigate the "crm stop" and see whether this creates the additional entry.
> >>>>
> >>>>
> >>>>
> >>>> Cheers,
> >>>> Bob Haxo
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>> On Thu, 2014-02-20 at 10:12 +1100, Andrew Beekhof wrote:
> >>>>> Have you been mixing pcs and crmsh again? :-)
> >>>>> The interesting part is how the dup got in there in the first place.
> >>>>>
> >>>>> Can you remove both settings and try to recreate that step?
> >>>>>
> >>>>> On 20 Feb 2014, at 9:39 am, Bob Haxo <bhaxo at sgi.com> wrote:
> >>>>>
> >>>>>> Chris, was easy to duplicate ... I thought that I had cleared
> >>>>>> the error, but that had not happened.
> >>>>>>
> >>>>>> Bob Haxo
> >>>>>>
> >>>>>> [root at mici-admin ~]# pcs resource disable virt
> >>>>>> [root at mici-admin ~]# pcs resource disable libvirtd-clone
> >>>>>> Error: Error performing operation: Invalid argument
> >>>>>> Multiple attributes match name=target-role
> >>>>>> Value: Started (id=libvirtd-clone-meta-target-role)
> >>>>>> Value: Started (id=libvirtd-clone-meta_attributes-target-role)
> >>>>>>
> >>>>>> [root at mici-admin ~]# pcs --debug resource disable libvirtd-clone
> >>>>>> Running: /usr/sbin/crm_resource -r libvirtd-clone -m -p target-role -v Stopped
> >>>>>> Return Value: 22
> >>>>>> --Debug Output Start--
> >>>>>> Error performing operation: Invalid argument
> >>>>>> Multiple attributes match name=target-role
> >>>>>> Value: Started (id=libvirtd-clone-meta-target-role)
> >>>>>> Value: Started (id=libvirtd-clone-meta_attributes-target-role)
> >>>>>>
> >>>>>> --Debug Output End--
> >>>>>>
> >>>>>> Error: Error performing operation: Invalid argument
> >>>>>> Multiple attributes match name=target-role
> >>>>>> Value: Started (id=libvirtd-clone-meta-target-role)
> >>>>>> Value: Started (id=libvirtd-clone-meta_attributes-target-role)
> >>>>>>
> >>>>>> [root at mici-admin ~]# pcs resource disable libvirtd-clone
> >>>>>> Error: Error performing operation: Invalid argument
> >>>>>> Multiple attributes match name=target-role
> >>>>>> Value: Started (id=libvirtd-clone-meta-target-role)
> >>>>>> Value: Started (id=libvirtd-clone-meta_attributes-target-role)
> >>>>>>
> >>>>>>
> >>>>>> On Wed, 2014-02-19 at 15:23 -0600, Chris Feist wrote:
> >>>>>>> On 02/19/2014 11:16 AM, Bob Haxo wrote:
> >>>>>>>> Encountered this error with pcs but not with crm. Looks like a regex
> >>>>>>>> error, with the existing regex grabbing all strings starting with the
> >>>>>>>> string "libvirtd-clone" ...
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> [root at mici-admin2 ~]# pcs resource disable libvirtd-clone
> >>>>>>>> Error: Error performing operation: Invalid argument
> >>>>>>>> Multiple attributes match name=target-role
> >>>>>>>> Value: Started (id=libvirtd-clone-meta-target-role)
> >>>>>>>> Value: Started (id=libvirtd-clone-meta_attributes-target-role)
> >>>>>>>>
> >>>>>>>> [root at mici-admin ~]# crm resource stop libvirtd-clone
> >>>>>>>>
> >>>>>>>> <<no error, success>>
> >>>>>>>>
> >>>>>>>> [root at mici-admin ~]# pcs resource enable libvirtd-clone
> >>>>>>>> Error: Error performing operation: Invalid argument
> >>>>>>>> Multiple attributes match name=target-role
> >>>>>>>> Value: Stopped (id=libvirtd-clone-meta-target-role)
> >>>>>>>> Value: Stopped (id=libvirtd-clone-meta_attributes-target-role)
> >>>>>>>>
> >>>>>>>> xml:
> >>>>>>>> <clone id="libvirtd-clone">
> >>>>>>>> <meta_attributes id="libvirtd-clone-meta">
> >>>>>>>> <nvpair id="libvirtd-clone-meta-target-role" name="target-role" value="Stopped"/>
> >>>>>>>> <meta_attributes id="libvirtd-clone-meta_attributes">
> >>>>>>>> <nvpair id="libvirtd-clone-meta_attributes-target-role" name="target-role" value="Stopped"/>
> >>>>>>>>
> >>>>>>>> Regards,
> >>>>>>>> Bob Haxo
> >>>>>>>
> >>>>>>> Can you send the output of the following command:
> >>>>>>> pcs --debug resource disable libvirtd-clone
> >>>>>>>
> >>>>>>> Thanks,
> >>>>>>> Chris
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> _______________________________________________
> >>>>>>>> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> >>>>>>>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> >>>>>>>>
> >>>>>>>> Project Home: http://www.clusterlabs.org
> >>>>>>>> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> >>>>>>>> Bugs: http://bugs.clusterlabs.org
> >>>>>>>>
> >>>>>>>
> >>>>>>
> >>>>>>
> >>>>>> _______________________________________________
> >>>>>> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> >>>>>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> >>>>>>
> >>>>>> Project Home: http://www.clusterlabs.org
> >>>>>> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> >>>>>> Bugs: http://bugs.clusterlabs.org
> >>>>>
> >>>>> _______________________________________________
> >>>>> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> >>>>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> >>>>>
> >>>>> Project Home: http://www.clusterlabs.org
> >>>>> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> >>>>> Bugs: http://bugs.clusterlabs.org
> >>>>
> >>>>
> >>>>
> >>>>
> >>>> _______________________________________________
> >>>> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> >>>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> >>>>
> >>>>
> >>>>
> >>>> Project Home: http://www.clusterlabs.org
> >>>> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> >>>> Bugs: http://bugs.clusterlabs.org
> >>
> >
>
More information about the Pacemaker
mailing list