[Pacemaker] Are you correct about the calculation of quorum when booth renews a ticket?

[Jiaju Zhang]

Hi Yuichi,

Thanks for the reporting! It should be a bug that learned_max has not
been reset after a successful learning process.

Thanks,
Jiaju

On Tue, 2012-12-25 at 11:28 +0900, Yuichi SEINO wrote:
> Hi Jiaju,
> 
> Even if booth lacks a quorum, booth may write a ticket in cib when
> booth renews a ticket. Are you correct about this work?
> 
> I try inserting the following debug code.
> 
> diff --git a/src/paxos.c b/src/paxos.c
> index c0ab5b9..53eadb5 100644
> --- a/src/paxos.c
> +++ b/src/paxos.c
> @@ -145,6 +145,7 @@ static int have_quorum(struct paxos_space *ps, int member)
>                         sum++;
>         }
> 
> +       log_debug("total %d, current %d", sum, member);
>         if (member * 2 > sum)
>                 return 1;
>         else
> 
> And, I got the following debug log when booth renews a ticket. And, I
> made 5 sites.
> ...
> booth-site[26305]: 2012/12/25_02:08:40 debug: learner response ...
> booth-site[26305]: 2012/12/25_02:08:40 debug: total 5, current 78
> booth-site[26305]: 2012/12/25_02:08:40 debug: enter lease_learned
> ...
> booth-site[26305]: 2012/12/25_02:08:40 debug: learner response ...
> booth-site[26305]: 2012/12/25_02:08:40 debug: total 5, current 79
> booth-site[26305]: 2012/12/25_02:08:40 debug: enter lease_learned
> 
> I think that the number of  vote is cumulative sum when booth renews a ticket.
> 
> Sincerely,
> Yuichi
> 
> --
> Yuichi SEINO
> METROSYSTEMS CORPORATION
> E-mail:seino.cluster2 at gmail.com