Description of problem: would be great to have an option for the cman init script which allows to call "cman_tool leave remove" to recalculate quorum when cman is stopped. looks like this is the only clean option to call "cman_tool leave remove". when you call it manually it doesn't work since a lot of processes rely on cman (fenced, ccsd, ...), actually you have to kill them manually in order to call "cman_tool leave remove". Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
add an option for LEAVE REMOVE of course :)
I think this would be a good thing to have for another reason, cman_tool leave remove doesn't seem to be tested at the moment and this might help that, particularly as it seems to be broken in alternate releases at the moment.
This request was evaluated by Red Hat Product Management for inclusion in a Red Hat Enterprise Linux maintenance release. Product Management has requested further review of this request by Red Hat Engineering, for potential inclusion in a Red Hat Enterprise Linux Update release for currently deployed products. This request is not yet committed for inclusion in an Update release.
It appears that the default behavior in RHEL5 is to call 'cman_tool leave remove' when calling the init script with the 'stop' argument. Is this the correct thing to do? According to the cman_tool man page, 'leave remove' should be done when the node is expected to be offline to an extended period of time. If having 'cman_tool leave remove' be the default operation is ok, then is there any need to add a new option? For example, we could add a 'stopremove' command to the cman init script, but is that desirable?
BZ #179627 made 'cman_tool leave remove' the default behavior when calling 'service cman stop'.
Personally I don't like having "cman_tool leave remove" as the default. Though if its been shipped like that it's probably hard to change it back.
Yes, I agree. It seems that the default should be to simply 'leave'. What are the risks associated with moving back to having the default case be to just call 'cman_tool leave'? It seems that this bug is asking for an option such that we can tell the init script that we want to do a 'leave remove' as opposed to a 'leave'. But since the default case is the do a 'leave remove', there is nothing to do for this BZ. However, if we go back to 'leave' being the default (which I think is correct), I think adding a 'stopremove' command to the init script makes sense. The way to init script is coded at the moment is weird. It checks $2 to see if it is defined (it never actually uses this parameter). If it is, then we do a 'cman_tool leave'. If it is not defined, we do a 'cman_tool leave remove' (default). The result is: % service cman stop => 'cman_tool leave remove' % service cman stop foo => 'cman_tool leave' where "foo" can be ANYTHING. Hrm. That doesn't seem right.
are you sure that quorum is recalculated when you stop cman? haven't tested that for a while. will do that tomorrow and provide further infos then.
(In reply to comment #9) > are you sure that quorum is recalculated when you stop cman? haven't tested that > for a while. will do that tomorrow and provide further infos then. Yes. I tested this recently, and according to the output from 'cman_tool status', quorum is recalculated after you for a 'service cman stop'. Test was done on RHEL 5.2.
Closing this since 'cman_tool leave remove' is the defaults behavior in RHEL5. This was changed with BZ 179627.
(In reply to comment #9) > are you sure that quorum is recalculated when you stop cman? haven't tested that > for a while. will do that tomorrow and provide further infos then. Also wanted to note that if you observe that 'service cman stop' does not result in quorum being recalculated, it could be a bug with 'cman_tool leave remove'. If that is the case, please note the version of cman you are using and create a new bug. Thanks!
ok, thanks, will let you know if I observe further issues.