FAILBACK_POLICY option

[Ramesh.K.R.]

Hi,

What is the best/recommanded setting for the FAILBACK_POLICY option ??
I think it is "MANUAL". But, i wanted the experts openion on the pro's & con's for this perticular option.

Thanks in advance,
Regards,
Ramesh.K.R.

[Manish Srivastava]

Hi,

The default policy is MANUAL. which means no attempt will be made to move the package back to it primary node when it is running on an alternate node. The alternate policy is AUTOMATIC which means the package will be moved back to its primary node whenever the primary node is capable of running the package.

MANUAL is better because you have to make sure the package will not fail again on the promary node once it it up.

manish

[Sridhar Bhaskarla]

Hi Ramesh,

FAILBACK_POLICY set to automatic is a dangerous option in my opinion. Consider the following scenario.

1. sysA and sysB are in a cluster with pkgA configured for sysA and pkgB for sysB.

2. cmcld is set to 1 so when a system is booting, it will automatically join the cluster.

3. FAILBACK_POLICY is set to automatic for pkgA.

4. sysA crashed due to a bad CPU. pkgA fails over to sysB.

5. sysA comes backup and joins the cluster. pkgA will be automatically moved to sysA.

6. Bad CPU shows up again on sysA. Step 4 and 5 will go in a continuous loop.

My preference is the default option which is MANUAL. This way, I can have sysA fixed before I manually move the package.

There may be cases where the load on the system is a concern whereas multiple downtimes is not and the application failover is really fast, you could go with AUTOMATIC option. But i find it very rare.


-Sri

[Sanjay_6]

Hi,

Check this link from hp doc site.

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B3936-90065/B3936-90065_top.html&con=/hpux/onlinedocs/B3936-90065/00/00/18-con.html&toc=/hpux/onlinedocs/B3936-90065/00/00/18-toc.html&searchterms=FAILBACK_POLICY&queryid=20040607-142119

The pros and con's are explained with pictures.

Hope this helps.

Regds

[Ashwani Kashyap]

MANUAL is the option I prefer bcause it allows you to troubleshoot the problem your are having without bothering about down time on the package .

[Stephen Doud]

The FAILBACK_POLICY setting depends on what you need the package to do in the event that it is running on the adoptive node when the primary server for the package comes back online.

If the adoptive server is not overburdened with workload, why not let it continue to operate the package until end-of-workday when it can be moved back to the primary server? This may be particularly advisable if the primary server has suffered an unexpected outage, suggesting a little monitoring to instill confidence before going back into service.

But if the adoptive server is not performing up to desired throughput because it was not expected to remain the major workhorse, one could argue that it is better to have the users wait out a short outage while the package moves back to the primary server where user requests will be handled with more responsiveness.

-sd-

[Rita C Workman]

FAILBACK Policy is strictly a decision that must be made internally.

So one can really only state that the software says the default option is Manual; and then state the option they choose to use.

For our decision....Manual. That way we would have the opportunity to check things and when we felt it was appropriate, then manually switch pkg back and ensure the failover has again been enabled.

Rgrds,
Rita

[Ramesh.K.R.]

Thanks for everyone, especially Sajay, the doc was really helpfull in stamping my openion on "MANUAL".

Ramesh.K.R.