raj14243 commented on code in PR #1926:
URL:
https://github.com/apache/cassandra-gocql-driver/pull/1926#discussion_r2835342224
##########
policies.go:
##########
@@ -455,35 +457,57 @@ func (t *tokenAwareHostPolicy) IsLocal(host *HostInfo)
bool {
}
func (t *tokenAwareHostPolicy) KeyspaceChanged(update KeyspaceUpdateEvent) {
- t.mu.Lock()
- defer t.mu.Unlock()
- meta := t.getMetadataForUpdate()
- t.updateReplicas(meta, update.Keyspace)
- t.metadata.Store(meta)
+ if update.Change != KeyspaceDropped {
+ t.updateReplicas(update.Keyspace)
+ }
}
-// updateReplicas updates replicas in clusterMeta.
-// It must be called with t.mu mutex locked.
+// updateReplicas updates replicas in clusterMeta for keyspace schema changes.
// meta must not be nil and it's replicas field will be updated.
-func (t *tokenAwareHostPolicy) updateReplicas(meta *clusterMeta, keyspace
string) {
- newReplicas := make(map[string]tokenRingReplicas, len(meta.replicas))
-
+func (t *tokenAwareHostPolicy) updateReplicas(keyspace string) {
Review Comment:
I chose to optimize for schema changes rather than startup time. We pay the
initialization cost once, but for future schema updates the number of keyspaces
that need changes will probably be low. In that case `updateReplicas` will be
called only a few times, whereas `updateAllReplicas` loops through all
keyspaces. I’m not married to either approach, since we compute once per
strategy, both should be fast, so I’ll follow your suggestion.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
