`cs_shadow` and `cs_commit` should not create changes every run. about puppet-corosync HOT 9 CLOSED

@hunner: is this still relevant? I'm trying to use the cs_shadow on a CentOS 6.4 box, and after fixing the mess caused by the SuSE crmsh package (see https://savannah.nongnu.org/bugs/?39013#comment0) I can easily list all shadow CIBs on the system with a simple crm cib list call.

To remove the lingering of the CIB on the system, it should be quite easy to delete the CIB once it has been committed in the cs_commit sync function, right?

//edit, just thinking about it, currently the instances function of cs_shadow always returns an empty array. This causes the resource to be created every Puppet run, even when no Pacemaker changes are needed at all, which is quite silly. By leaving the shadow CIB on the system and implementing the instances function using crm cib list you could avoid this. Not sure if the remaining data in the shadow CIB could cause an issue when the cluster is out of sync with the shadow file and Puppet makes a change, maybe you'd need a way to reset the shadow CIB on changes first.

from puppet-corosync.

@kbon

Not sure if the remaining data in the shadow CIB could cause an issue when the cluster is out of sync with the shadow file and Puppet makes a change, maybe you'd need a way to reset the shadow CIB on changes first.

That will definitely cause a problem. When the shadow CIB is created, it is initially a mirror of the live CIB. If they get out of sync, then the shadow is committed, it will have the effect of reverting the live CIB to the state it was in when the shadow was created, plus whatever changes were made to the shadow. Probably not what's desired.

Currently this is avoided because cs_shadow deletes, then recreates, the shadow CIB on each run.

from puppet-corosync.

@bitglue: indeed, that's also what I experienced. I've made some code changes to properly use the shadow file and only when needed, even got rid of cs_commit altogether. Stay tuned for a commit & PR!

from puppet-corosync.

Is this still an issue?

from puppet-corosync.

Yes and a fix is coming

from puppet-corosync.

Can you link to the relevant PR?

from puppet-corosync.

#162

from puppet-corosync.

@roidelapluie what's the current pr for this particular issue?

from puppet-corosync.

Fixed in #209

from puppet-corosync.