This repository was archived by the owner on Oct 9, 2023. It is now read-only.
Bind Cluster sessions to single server nodes, and add naive load balancing #183
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
alexjpwalker
commented
Feb 11, 2021
| self.options = options | ||
|
|
||
| def run(self, replica: ReplicaInfo.Replica): | ||
| return SessionClusterRPC(self.client, replica.address(), self.database, self.session_type, self.options) |
Member
Author
There was a problem hiding this comment.
FailsafeTask is a new class that holds the failover/retry logic. It is an abstract base class, requiring that run is implemented, where run represents a unit of work that can fail and should be retried / failed over.
| self.cluster_session.core_session.close() | ||
| self.cluster_session.core_client = self.cluster_session.cluster_client.core_client(replica.address()) | ||
| self.cluster_session.core_session = self.cluster_session.core_client.session(self.database, self.cluster_session.session_type(), self.options) | ||
| return self.cluster_session.core_session.transaction(self.transaction_type, self.options) |
Member
Author
There was a problem hiding this comment.
In FailsafeTask, run is called on the first run, and rerun is called on every subsequent run, if the first run fails.
In this case, if the first run fails, we need to tell the Cluster session to:
- close its Core session
- switch to a different Core client (linked to a different server node)
- try to open a new Core session in the new client
- try to open the transaction using this Core session, instead of the original one.
This allows us to fail-over the opening of Transactions, while enforcing that each Cluster session has, at any one time, precisely one Core session associated with it.
vmax
reviewed
Feb 12, 2021
vmax
approved these changes
Feb 12, 2021
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What is the goal of this PR?
Previously, each Cluster session would lazily create Core sessions on-demand as Transactions were opened. This delegated too much responsibility to the Transaction. Now, each Cluster session is bound to a single server node, which may be a primary or a secondary node depending on the Options passed to the Session.
We also select secondary nodes in a fairer, more balanced way; the server indicates which node is its first preference for secondary sessions, and the client tries to obey the server's decision, with fallback to backup nodes only in the event of a failure.
What are the changes implemented in this PR?
read_any_replicaOption passed to the Sessionread_any_replicaistrue, the client will first select nodes that the server has indicated are "preferred secondary", with fallback to backup nodes only in the event of a failure