We continue discussing the ZooKeeper. It is a co-ordination service with elements from group messaging, shared registers and distributed lock services. It provides a interface that guarantees wait-free property and FIFO execution of requests from each client. Requests across all clients are also linearized.
Both the linearizable writes and the FIFO client order form the guarantees for ZooKeeper. These two guarantees also interact. ZooKeeper handles both by designating a path as the "ready" ZNode. The other clients may make changes only when that ZNode exists.
There are a few caveats with these guarantees.
If a process sees the ready exists before the new leader starts to make a change, then starts reading the configuration while the change is in progress, then it will see an inconsistent state. This is solved with the ordering guarantees of the notifications where the notification happens before the client sees the new state.
Ideas for source code visualization tool: https://1drv.ms/w/s!Ashlm-
Both the linearizable writes and the FIFO client order form the guarantees for ZooKeeper. These two guarantees also interact. ZooKeeper handles both by designating a path as the "ready" ZNode. The other clients may make changes only when that ZNode exists.
There are a few caveats with these guarantees.
If a process sees the ready exists before the new leader starts to make a change, then starts reading the configuration while the change is in progress, then it will see an inconsistent state. This is solved with the ordering guarantees of the notifications where the notification happens before the client sees the new state.
Ideas for source code visualization tool: https://1drv.ms/w/s!Ashlm-
No comments:
Post a Comment