Homework 3 forum

Clarifications of PaxosSequenceID

 
Avatar Bastien Wermeille
Clarifications of PaxosSequenceID
par Bastien Wermeille, mardi, 1 décembre 2020, 17:38
 

Hello,

I have implemented my paxos and all my tests are working except for some with the binary gossiper. I have however some warnings when running with the binary gossiper :

- paxosSeq id is lower address=127.0.0.1:36508 type=blockchain

- ignoring message {PaxosSeqID:0 IDp:0 IDa:-1 Value:{BlockNumber:0 PreviousHash:[] Metahash:[] Filename:}} because we are not preparing: proposing [36mID=[0m127.0.0.1:58798 [36mtype=

- TLC received for a past block index 0. Ours is 1 [36maddress=[0m127.0.0.1:58798

Could you precise what exactly PaxosSeqID is ? I thought that it was a equal to the block number so starting from 0 and grwing up at each step of new consensus but I'm starting to doubt my understanding.

Regarding the promise packet type, what value should be set to IDa when we accept a propose message ? In the handout it's written "IDa is the highest ID the acceptor saw", so shouldit be -1 or the value we just want to accept for example 0.

Any help and idea is most welcome.

Thanks for your reply.

Avatar Pasindu Nivanthaka Tennage
Re: Clarifications of PaxosSequenceID
par Pasindu Nivanthaka Tennage, vendredi, 4 décembre 2020, 15:26
 

Hello,

What exactly isPaxosSeqID?
 it is equal to the block number so starting from 0 and growing up at each step of new consensus: your answer is correct

What value should be set to IDa when we accept a propose message ?

If the node hadn’t previously accepted any value, it sets IDa = -1. Otherwise IDa is the ID of the proposal that the acceptor already accepted for this Paxos box. the value IDa doesn’t strictly matter. If the block in the promise is set to types.Block{} , it means the node hadn’t previously accepted any value, so IDa is irrelevant. Otherwise, IDa is relevant. There is already a post about this value for block in the promise https://moodle.epfl.ch/mod/forum/discuss.php?d=48929


Regards

Avatar Bastien Wermeille
Re: Clarifications of PaxosSequenceID
par Bastien Wermeille, samedi, 5 décembre 2020, 13:08
 
Thanks for your reply.