zbentley commented on issue #12551: URL: https://github.com/apache/pulsar/issues/12551#issuecomment-955749384
> > This is probably more important: all operations related to deletion/creation of topics/namespaces/tenants conclude before the management API returns success codes to the user. > > The operations are always acked after the writes are completed. What typically happens is that the 2nd request goes to a different broker which has not yet received a cache invalidation signal (eg: zk watch). @merlimat that is good info. It indicates a partial workaround while these conditions are addressed; I can force clients doing "DDL" type work to always connect to a single broker. However, that still will likely hit variations on these issues; since a given container entity (say a partitioned topic, or a namespace) can have sub-entities resident on multiple brokers, attempting to perform CRUD operations on entities on those other brokers (e.g. by having my client "chase" the 307 responses returned from the primary DDL broker when I attempt to do an action on something that's not mastered on that broker) will still encounter issues due to the other brokers' metadata caches not yet being updated. -- 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]
