[Drbd-dev] Summary (=Idems for 0.8 Roadmap)

[Drbd-dev] Summary (=Idems for 0.8 Roadmap)

[Philipp Reisner]

Hi,

New configuration options:

  primary-after-sb  (SB = splitbrain) = 
   older		The node that was primary before split brain (current behaviour)
   younger		The node that became primary during split brain 
   most-mod		The node that modified more of it's data during the SB.
   wait			None, wait for operator's decission. 
   node-"nodename"	The node prefered by the user.[suggested by Helmut]

  wrong-gencnt (secondary-with-higher-genctn) =
   refuse		(Current behaviour) I think the secondary should drop its net-conf.
   update-data		Make the joining node sync-target.

  secondary-after-sb = 
   keep-data		The secondray drops its net-config.
   update-data		Make the looser sync-target (Current behaviour)

What do you think ?

-Philipp
-- 
: Dipl-Ing Philipp Reisner                      Tel +43-1-8178292-50 :
: LINBIT Information Technologies GmbH          Fax +43-1-8178292-82 :
: Schönbrunnerstr 244, 1120 Vienna, Austria    http://www.linbit.com :

Re: [Drbd-dev] Summary (=Idems for 0.8 Roadmap)

[Lars Ellenberg]

/ 2004-08-25 12:07:20 +0200
\ Philipp Reisner:
> Hi,
> 
> New configuration options:
> 
>   primary-after-sb  (SB = splitbrain) = 
>    older	The node that was primary before split brain (current behaviour)
>    younger	The node that became primary during split brain 
>    most-mod	The node that modified more of it's data during the SB.
>    wait		None, wait for operator's decission. 

would this block current activity on the Primary/both nodes?
what if _both_ are still Primary?

>    node-"nodename"	The node prefered by the user.[suggested by Helmut]

>   wrong-gencnt (secondary-with-higher-genctn) =
>    refuse	(Current behaviour)
>          	I think the secondary should drop its net-conf.

currently the Primary goes StandAlone.
you suggest the Secondary goes StandAlone, too,
                       or goes StandAlone instead?

>    update-data	Make the joining node sync-target.

so, in a peer to peer scenario, who joins whom?
this of course means make the Secondary with the higher gencounts
sync-target in the presence of a running Primary.

this is a delicate decision... do we really want this to be automated?
maybe that could become at least an undocumented feature?

additional questions:
who would adopt the gencounts of the other?
do we want to preserve strict non-decreasing monotonicity?



>   secondary-after-sb = 
>    keep-data		The secondray drops its net-config.
>    update-data	Make the looser sync-target (Current behaviour)

um. I don't see this at all?
what scenario is this?

	lge