findif.sh returns OCF_ERR_CONFIGURED, should it occasionaly return OCF_ERR_INSTALLED instead? about resource-agents

[dmuhamedagic]

@kskmori: issue opened as promissed :)

[kskmori]

Regarding to IPaddr2 at least, I think that OCF_ERR_INSTALLED does not make sense.

I don't think that it is a realistic situation that an user specified, say, eth2 "correctly" but actually the user "forgot" to install the third NIC device :) It must be more likely the user mistook the interface name. Anyway it is impossible to distinguish between them essentially.

Also, unlike apache or pgsql, IPaddr2 does not require an extra software package that the RA is monitoring to.

So I think it is sane that IPaddr2/findif.sh only returns OCF_ERR_CONFIGURED in such parameter checking.
The previous version of the findif binary returns OCF_ERR_CONFIGURED only, too.

By looking at the current code, IPaddr2 returns OCF_ERR_INSTALLED only when the platform is not Linux:)

[dmuhamedagic]

Thanks for the explanation. I'm not sure under which circumstances the NIC can disappear. For instance, are there NIC failures which can cause that it disappears from the system? If so, then returning OCF_ERR_CONFIGURED would prevent pacemaker from starting the resource on any node in the cluster. I'm still not sure if that can happen at all, but it's worth considering.

[kskmori]

OK, now I see your point. I will reconsider it...

[kskmori]

To the best of my knowledge, NIC can disappear when 1) ifdown 2) unplug a USB NIC etc. and it should not disappear when a "normal" failure, but nobody can guarantee it...

How about doing the findif.sh check only on start and validate-all since the purpose of the check is just a sanity check of the parameters specified by an user and not necessarily doing it on every monitor?

By the way, in the case of #320 a fail over was exercised with the scenario below:

• an IPaddr2 detected a transient failure and returned OCF_ERR_CONFIGURED
• Pacemaker tried to stop the upper resources which take some time
• meanwhile, the failed IPaddr2 was monitored again and now succeeded.
• Pacemaker started to move the resources because the fail-count is not 0. No record of OCF_ERR_CONFIGURED any more.

How lucky I was:)

[dmuhamedagic]

On Fri, Oct 11, 2013 at 01:30:09AM -0700, Keisuke MORI wrote:

How about doing the findif.sh check only on start and validate-all since the purpose of the check is just a sanity check of the parameters specified by an user and not necessarily doing it on every monitor?

Agreed. The less code we need to execute in monitor the better.

By the way, in the case of #320 a fail over was exercised with the scenario below:

• an IPaddr2 detected a transient failure and returned OCF_ERR_CONFIGURED

Oops.

• Pacemaker tried to stop the upper resources which take some time
• meanwhile, the failed IPaddr2 was monitored again and now succeeded.
• Pacemaker started to move the resources because the fail-count is not 0. No record of OCF_ERR_CONFIGURED any more.

How lucky I was:)

Indeed :)