[ClusterLabs] big trouble with a DRBD resource

Kristoffer Grönlund kgronlund at suse.com
Thu Aug 10 16:21:48 EDT 2017

"Lentes, Bernd" <bernd.lentes at helmholtz-muenchen.de> writes:

> In both cases i'm inside crmsh.
> The difference is that i always enter the complete command from the highest level of crm. This has the advantage that i can execute any command from the history directly.
> And this has a kind of autocommit.
> If i would enter a lower level, then my history is less useless. I always have to go to the respective level before executing the command from the history.
> But then i have to commit.
> Am i the only one who does it like this ? Nobody stumbled across this ?
> I always wondered about my ineffective commit, but never got the idea that such a small difference is the reason.

You are right, this is a quirk of crmsh: Each level has its own "state",
and exiting the level triggers a commit. Running a command like
"configure primitive ..." results internally in three movements;

* enter the configure level: This fetches the CIB and checks that it is writable
* create the primitive: This updates the internal copy of the CIB
* exit the configure level: This creates, verifies and applies a patch to the CIB

I can't speak for others, but somehow this has never caused me problems
as far as I can remember. Either I have been using it interactively from
within the configure section, or I have been running commands from
bash. I can't recall if that's because I was told at some point or if it
was made clear in the documentation somewhere.


// Kristoffer Grönlund
// kgronlund at suse.com

More information about the Users mailing list