0

I read the paxos can be used to keep replications in sync by syncing operation logs to replications.

To my understanding each paxos instance will define an log id with the operation

Ideally the log will be like this on each nodes, therefore they apply the operation log in the order of log id to keep the data consistent with each other

  1. Delete ....
  2. Add ...
  3. Remove ...

To my understanding the log id is not increased one by one , so the log should like this :

1.Delete ....

3.Add ...

5.Remove ...

I guess my questions is how it is work practically if one node is down, during the paxos instance?

Because this node will miss one log entry, how does this node knows it missed an entry after it recovered ?

According to wiki

https://en.wikipedia.org/wiki/Paxos_(computer_science)#Multi-Paxos

Multiple-paxos seems implemented in way of make sure LOG ID always increase with no gap.

That's said, to my understanding, each replicate still need to periodically talk to some to make sure the are on the latest change.

DCY
  • 1
  • 2
  • Why would you not make the log id increase by one? In all practical implementations I know of they must only increase by one. – Michael Deardeuff Aug 09 '15 at 04:45
  • Yes, you are right, I just revisit the paper and wiki, it looks the log id is increased by one. – DCY Aug 10 '15 at 02:20

0 Answers0