IGNITE-19494 Correctly stop a replica#2142
Conversation
|
|
||
| //TODO: IGNITE-15568 endOfBatch should be set to true to prevent caching tasks until IGNITE-15568 has fixed. | ||
| handler.onEvent(event, sequence, true); | ||
| handler.onEvent(event, sequence, subscribers.size() > 1 ? true : endOfBatch); |
There was a problem hiding this comment.
Not intended, but I decided to leave it after commit. Because this allows to do slightly better of disruptor for Meta storage and Cluster management group and I don't want to create dedicate ticket for this small change.
There was a problem hiding this comment.
Sounds reasonable, however I had to ask. How did you understand that it works better after given change?
There was a problem hiding this comment.
We don't have a test that will be able to demonstrate that. I assume, that update is not doing worse at least.
Theoretically, RAFT log may fsync less often in the circumstance when several RAFT commands are executing simultaniously.
There was a problem hiding this comment.
I don't think that this change is reasonable, because we have a plan to port disruptor changes in a upcoming sprint, potentially this fix will live a few weeks
There was a problem hiding this comment.
We will work a little better just now. Today, the change worth nothing for us.
modules/replicator/src/main/java/org/apache/ignite/internal/replicator/ReplicaManager.java
Show resolved
Hide resolved
modules/replicator/src/main/java/org/apache/ignite/internal/replicator/ReplicaManager.java
Show resolved
Hide resolved
.../src/main/java/org/apache/ignite/internal/replicator/exception/ReplicaStoppingException.java
Show resolved
Hide resolved
3 participants
https://issues.apache.org/jira/browse/IGNITE-19494