[Linux-HA] pingd behaviour

Terry L. Inzauro tinzauro at ha-solutions.net
Fri Oct 26 11:28:10 MDT 2007


Andrew Beekhof wrote:
> On 10/25/07, Terry L. Inzauro <tinzauro at ha-solutions.net> wrote:
>> Andrew Beekhof wrote:
>>> On 10/23/07, Terry L. Inzauro <tinzauro at ha-solutions.net> wrote:
>>>> list,
>>>>
>>>> i think i have pingd working properly.
>>>>
>>>> /etc/ha.d/ha.cf
>>>>         apiauth ping gid=root uid=root
>>>>         respawn root /usr/lib/heartbeat/pingd -m 1000 -d 5s -a default_ping_set
>>> here use tell pingd to define "default_ping_set"
>>>
>>>> cib.xml locational constraint:
>>>> <constraints>
>>>>        <rsc_location id="afp_cl_vpn_loc" rsc="afp_cl_vpn">
>>>>          <rule id="afp_cl_vpn_pref_1" score="100">
>>>>            <expression id="afp_cl_vpn_loc_attr_1" attribute="#uname" operation="eq" value="clvpn01"/>
>>>>          </rule>
>>>>          <rule id="afp_cl_vpn_loc_pingd_rule" score_attribute="afp_cl_vpn">
>>>>            <expression id="afp_cl_vpn_loc_pingd_rule_defined" attribute="pingd" operation="defined"/>
>>> but here you tell the PE to look for "pingd"
>>>
>>> s/pingd/default_ping_set/
>>>
>>>>          </rule>
>>>>        </rsc_location>
>>>> </constraints>
>>>>
>>>>
>>>> my question is in regards to resource fail back.  how does one get the resource to move back to the
>>>> primary node in the event that the resource moves from its primary cluster node to the secondary
>>>> node as a result of its pingd score?
>>> it depends on how default-resource-stickiness is set
>>> if its > 0, then you need to move it manually with crm_resource -M
>>>
>>>> is this a manual administrtive process or 'should' hb take care
>>>> of that for you?
>>>>
>>>>
>>>> regards,
>>>>
>>>>
>>>> _Terry
>>>>
>>
>> ok, after resolving my mistakes in naming, i began to test.  the way i understand this to work is
>> that the resource will
>> be placed on the node with the highest score.  i'm simulating network failure by placing black hole
>> routes on the node
>> i wish to test (route add -host 172.16.1.2 gw 127.0.0.1). i would 'assume' that the resource would
>> be moved to the
>> node with the best connectivity aka score...it does, but it also offlines the node that i added the
>> black hole route to.
> 
> in what way did the node go offline?
> (in heartbeat, the crm, physically?)
> 
> if just in the crm, then please upgrade.  there was a bug about this
> some time ago.
> 
>> how does one reverse the effects of the off lining, short of restarting heartbeat of both nodes?
>>
>> is that what ping is supposed to do? if so, what if i had other resources configured to run on that
>> node?

yes, the node goes off line in crm, but heartbeat remains running but hangs when an attempting to
shut down.

i was hoping you weren't going to say that.  i'm using debian etch(stable) and i like keeping the
packages as close to their release cycle as possible.

clvpn01:/var/lib/heartbeat/crm# /usr/lib/heartbeat/heartbeat -V
2.0.7

can you recommend which version would be the most stable/recommended? ;) in sid, the latest one is:
http://packages.debian.org/sid/heartbeat/i386/download





More information about the Linux-HA mailing list