Vladimir Blagojevic wrote:
On 5/28/09 4:56 PM, Mircea Markus wrote:
> e it a try as well! I suggest we move to 2.8 sooner than later.
> Currently, with tcp and 2.7 I get 0 failures and with 2.8 I get 9
> faliures.
> That's something!
> The way I see it we are kind of trapped: on one hand we cannot fix
> NBST without a JGroups fix, and JGroups fix can only be performed on
> 2.8, and we cannot go live with it.
> When is 2.8 scheduled for going GA? Vladimir, the fix you intend to
> implement still binds us to FLUSH? If yes, we can also consider
> another fix to the issue (to drop FLUSH) and we won't depend upon
> JGroups release this way.
Mircea, if you have a solution that does not depend on FLUSH I think
that is the best way to go. Eventually, you'll have to drop FLUSH
anyway, at least that's what Manik wanted. I'll see with Bela to
augment FLUSH to cache those in-between messages anyway.
One solution I had in mind
is:
original view {A,B,C}
- A continuously broadcast messages
- D starts joining
- A receives the new view: {A,B,C,D}
- A multicasts to {A,B,C}
- D finishes the join (view install,state transfer) and sends and
broadcasts an JoinCompleteMessage
- A receives JoinCompleteMessage, and from now on broadcasts messages
Regards,
Vladimir
P.S Do a run with 2.8.0Alpha3 and see if it works for you as well.