[Sheepdog] partition recovery algorithm
boubouzzzz at orange.fr
Wed Nov 4 12:59:54 CET 2009
Chris Webb a écrit :
> MORITA Kazutaka <morita.kazutaka at lab.ntt.co.jp> writes:
>> Did you mention about a network partition problem (split-brain)?
>> Sheepdog does not have the tolerance for network partition
>> in current implementation.
>> We think of using a majority voting algorithm to deal with this problem,
>> but details have not been discussed enough.
> I think this is one of the areas where porting to use a cluster engine like
> corosync instead of jgroups may help you, quite apart from getting rid of
> the jvm dependency. Corosync already has a well-tested quorum service, which
> can be used to protect against network partition issues.
dog is nice coded in java.
why recoding it ? performance ?
nice to go from langage to another, no ?
and java isn't clean ?
ok jvm is heavy, that a bad point.
but many projects include already a jvm.
an other point, I never have several servers with info -t dog, when the
others write nice 0 and 4000 nice in directory. and VM seam to write in
only one server.
I'll check about that, but not priority
More information about the sheepdog