[Pacemaker] Time-based location constraint not working in HB 2.1.4?

[Andrew Beekhof]

On Dec 12, 2008, at 1:20 PM, Andreas Kurz wrote:

> Andrew Beekhof wrote:
>> On Fri, Dec 12, 2008 at 12:12, Andreas Kurz  
>> <andreas.kurz at linbit.com> wrote:
>>> Andrew Beekhof wrote:
>>>> On Dec 12, 2008, at 10:32 AM, Andreas Kurz wrote:
>>>>
>>>>> Andrew Beekhof wrote:
>>>>>> On Dec 11, 2008, at 1:53 PM, Andreas Kurz wrote:
>>>>>>
>>>>>>> Andrew Beekhof wrote:
>>>>>>>> Have you read the section on "Ensuring Time Based Rules Take  
>>>>>>>> Effect" ?
>>>>>>> Yes ... but in the "Configuration Explained" document for  
>>>>>>> Pacemaker 0.6
>>>>>>> this section is (nearly) empty.
>>>>>>>
>>>>>>> In the latest version for Pacemaker 1.0.1 I found the hint  
>>>>>>> about the
>>>>>>> "cluster-recheck-interval" ... but it has no effect on my  
>>>>>>> config ...
>>>>>>> too
>>>>>>> old Pacemaker version? .. a "ptest metadata" gives no  
>>>>>>> description for
>>>>>>> this attribute.
>>>>>> crmd metadata (its a crmd option)
>>>>> ahh ... good to know ;-)
>>>>>
>>>>>> try with underscores instead of the dashes
>>>>> hmm ... has no effect:
>>>>>
>>>>> # crm_attribute -G -n cluster_recheck_interval
>>>>> name=cluster_recheck_interval value=3min
>>>> did you restart the node?  its only read once - at startup.
>>>> otherwise it could be a bug in 2.1.4
>>> huh ... no, I did not restart the node ... are all these crmd  
>>> options
>>> only read at startup?
>>
>> correct - i should probably add the ability for admin's to initiate a
>> refresh.  care to create a bug for that?
>
> Bug 2015
>
>>
>>> Is a restart of heartbeat on both nodes needed ...
>>> or only on the current DC?
>>
>> Essentially everything.  Since restarting just the DC will result in
>> another node being the DC.
>
> Out of curiosity: why is the cluster_recheck_interval not enabled by
> default with a value like 1min or 5min?

Time-based rules aren't that common yet.