Agent Failover - use always DEFAULT EM, if it's available

[pivst01]

Hi All,

 

is there please any possibility to force the Agent to always (re)connect to the DEFAULT EM, when it's available

 

even in case it was unavailable, BACKUP EM took it's place, but after some time the DEFAULT EM comes up again? so it should be use..

 

1.

introscope.agent.enterprisemanager.connectionorder=DEFAULT,BACKUP

introscope.agent.enterprisemanager.failbackRetryIntervalInSeconds=30

DEFAULT & BACKUP EM are in different cluster, so i think loafbalancing.xml shouldnt be used as the agent shouldnt be loadbalanced, the agent should use the BACKUP EM just in some exceptional case.

2.

DEFAULT EM goes down and the agent connects to the BACKUP EM

3.

DEFAULT EM comes up and it should be used

 

i tried it, but the agent kept connecting to the BACKUP EM even the DEFAULT EM came up and were available again.

 

any idea?

 

thank you,

stefan

[Hallett German]

Hi Stefan:

   I asked internally for someone to respond

 

Thanks

Hal German

[Matthew Muskaloon]

Hi Stefan,

 

What version of the Agent and EM's is this happening with?

 

So that I am clear on this, when you have both Default and Backup EM's running, does it go to Default first?

If Default is down, does it go to Backup?

 

Basically if both Default and Backup are running and accepting communications, barring any clamps or anything else, in other words, running good, Default should be the first one to go to.  If Default is not available when the Agent starts or gets disconnected, Backup should get the Agent.

 

Thanks,
Matt

[pivst01]

Hi Matt,

 

yes, i thought so as well that it should work and will work as you describe = default is used always when it's available.however it doesnt switch back from backup.

 

it's v10.1 (both EM and the agent too).

 

any other idea or what to test/try is appreciated .

 

thank you,

stefan

[Gustavo Strabelli]

Hey Stefan,

 

I Had a similar situation with one customer and I could handle by enabling AgentFailback:

 

1- Set Collectors order and Collectors communications

2- Uncomment Failback threshold:

#introscope.agent.enterprisemanager.failbackRetryIntervalInSeconds=120

 

Restart the Agent and see if it helps.

 

Gustavo.

[pivst01]

Hi Gustavo,

 

both the order and the Failback interval is/was there:

 

introscope.agent.enterprisemanager.connectionorder=DEFAULT,BACKUP

introscope.agent.enterprisemanager.failbackRetryIntervalInSeconds=30

 

thank you,

stefan

[Hallett German]

Hi Stefan:

Please let us know after Sergio's response if there are any outstanding questions

Thanks

Hal German

[pivst01]

Hi Hal,

 

Sergio described different scenario with MoM that is not present in this environment, so im still waiting for the answer.

 

thank you,

stefan

[Sergio Morales Correa]

Hi Stefan,

I belive you are looking for a way to configure the Enterprise Manager Failback mechanism in the Loadbalancing.xml

First, the agent assignment is controlled by the MOM LB xml, you need to update the MOM_HOME\config\loadbalancing.xml as below, for example:

 

<agent-collector name="prod">

<agent-specifier>.*\|MQ.*\|.*</agent-specifier>

<include>

<collector host="DEFAULT" port="5001" latched="true" />

<collector host="BACKUP" port="5001" />

</include>

</agent-collector>

 

In the above case scenario, MQAgent will reconnect to the primary collector during the next load rebalancing which occurs by default every 10 min (introscope.enterprisemanager.loadbalancing.interval=600)

 

Best Regards,

Sergio

[pivst01]

Hi Sergio,

 

as i wrote in the description "DEFAULT & BACKUP EM are in different cluster, so i think loadbalancing.xml shouldnt be used as the agent shouldnt be loadbalanced, the agent should use the BACKUP EM just in some exceptional case."

 

so there is no MOM.

 

if you could advice how to configure just a basic primary and backup collector for an agent that uses always the primary if it's available (without MoM), i'd appreciate it.

 

thank you,

stefan

[Sergio Morales Correa]

Hi Stefan,

introscope.agent.enterprisemanager.failbackRetryIntervalInSeconds property applies only to agent attempting a connection. Once the agent acquires a connection, unless it is specifically redirected or EM goes down, it does not switch EMs.

About "how to configure just a basic primary and backup collector for an agent that uses always the primary if it's available"

Fallbacks and locking are within the cluster. Two clusters don't communicate between each other, except with "Enterprise Team Center" of course, which doesn't deal with agents or loadbalacning

I think this specifc case is not supported "Fallbacks and latching"

Regards,

Sergio

[pivst01]

Hi Sergio,

 

thank you for your answer.

 

and if it's in cluster and MOM is used as well, is there any possibility to configure MOM to force the agents to ALWAYS connect to the primary collector if it's available, even the primary is overloaded?

 

i tried it in the past and it just although it was configured as you describe it =

<collector host="DEFAULT" port="5001" latched="true" />

<collector host="BACKUP" port="5001" />

, but it spread the agents between the primary and backup to spread the performance load.

 

thank you,

stefan