>>>> On another thread () I've been trying to figure out how
>>>> to get pingd to play nicely with hb_gui. One of the
>>>> difficulties of this task is understanding
>>>> what kind of information pingd provides. Judging from
the
>>>> examples on the pingd web page, the possible return
>>>> values are something like
>>>>    defined
>>>>    not_defined
>>>>    <value>

>>> There are not exactly return values. The pingd program,
>>> which is either managed by the pingd RA or using the
>>> ha.cf respawn directive, just creates an attribute in
>>> the status section of the CIB. The value of that
>>> attribute depends on the number of nodes it was able to
>>> ping (n*multiplier).

>> I cannot find anything that looks like a score in the
>> status section given by "cibadmin -Ql". What should that
>> look like? And again, what does defined or undefined
mean?

> Look for "pingd". defined means that the attribute is
> present.

>> Curiously, I do see two sets of fail counts that
>> should be getting updated, but they are not. Since my
>> network cable on one computer is unplugged, one of the
>> fail counts should be going up, but it is not changing.

>>>> What do these mean? More importantly what does <value>
>>>> mean? The first two cannot be used with the gui because
>>>> it will not accept the "defined" or "not_defined"
>>>> operator without also entering a value. And
>>>> it seems that if you enter a value (e.g. " " or "true")
>>>> in the gui, heartbeat does not like it.

>>> Hmm, not an expert on GUI, but shouldn't you here enter
>>> the attribute name for the <value> (typically pingd).

> This is wrong. I expected the gui to support
score_attribute, but
> it doesn't.

>> Gui expects "attribute", "operation", and "value". I used
>> "pingd" for the attribute and have been changing the
>> operation, but with no luck.

>>>> So then we're left with trying to use <value> for a
>>>> failover trigger. I'm trying to figure out how to
combine
>>>> a location constraint with a pingd clone monitor to get
>>>> things to failover. Possible on_fail actions are
>>>>    ignore
>>>>    block
>>>>    stop
>>>>    restart
>>>>    fence
>>>> It's a little unclear what these mean in the context of
>>>> a clone and I'm not sure which one to use along with a
>>>> pingd location constraint based on
>>>> <value>

>>> The pingd clone is not important, it just serves to
>>> manage the pingd program. The pingd program does all
>>> the work. Don't reference that cloned resource in
>>> constraints. Just reference the attribute which pingd
>>> inserted in the status section (using score_attribute
>>> instead of score).

>> Gui doesn't support score_attribute, so I have been
trying
>> to get things to work with score.

>>> I suppose that you checked http://linux-ha.org/pingd

>> More times than I can count. Regrettably, I cannot use
>> the examples in that document because they have things
>> that simply are not supported by gui.

> True, score_attribute is not supported. But you should be
able to
> specify all other constraints, such as this one:

> <rsc_location id="my_resource:connected"
rsc="my_resource">
>   <rule id="my_resource:connected:rule" score="-INFINITY"
> boolean_op="or">
>   <expression id="my_resource:connected:expr:undefined"
>     attribute="pingd" operation="not_defined"/>
>   <expression id="my_resource:connected:expr:zero"
>     attribute="pingd" operation="lte" value="0"/>
>   </rule>
> </rsc_location>
>
> That constraint says: don't run my_resource on a node for
which
> pingd attribute is not defined or its value is not greater
than
> zero. Did you have problems specifying this one?

Yes, I did have problems specifying this one in the GUI.
There are two issues with the above configuration:
  (1) boolean_op is not supported
  (2) attributes without values (like
      defined or undefined) not supported
There is another potential way to attack this. Which is to
try to configure manually/with cib tools and then monitor/
administer with the GUI. Most of my experiments doing
it that way have failed, but it could be because I am
misconfiguring the cib.xml wrong in the first place.

I even took it one step further and started looking at
the code in the GUI. It looks like it's 2493 lines of
mixture of python, Glade, and GTK. 2493 lines.
Regrettably, I am not fluent in any of these,
but I'd like to figure out if there is a quick way to hack
in support for some of these features.


_______________________________________________
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