Matteo Merli

@flowchartsman In your example there would be no need to keep the message ids, you could use a Reader and do SeekByTime(): ``` // Reset the subscription associated with this...

> Error: Expected nil, but got: unable to subscribe to topic=persistent://public/default/my-topic-262453276-9: server error: TopicNotFound: Topic does not exist @flowchartsman I think it is related to the change. It looks like...

@billowqiu The default action is to block the `Send()` when the quota is full. You can change the behavior in the namespace (or topic) policies: https://pulsar.apache.org/docs/en/admin-api-namespaces/#configure-backlog-quota-policies

I'm not sure I understand why we need to have 2 different modes here. Do both methods provide the same guarantees or not?

I still cannot see a case where flushing the entry location index (ledgerId, entryId, offset) becomes the bottleneck, compared to: 1. Journal 2. Entry log flush (where all the data...

> In order to speed up the flush, we only need to ensure that the EntryLocationIndex is successfully written under normal circumstances, regardless of machine downtime. I'm not sure I...

@Shoothzj Can you also check the number of allocations? Using `-prof gc`. I remember in earlier versions of datasketches (after 0.8.3) they had introduced some regression that added lot of...

@Shoothzj Just did a quick test. There are 4 bytes per each recorded sample with the new Datasketches. It would be good to understand why that is the case and...

The restart is caused by the auto-recovery component of the bookies. In general, it is better to run the auto-recovery as a separate service (it's completely stateless), rather than as...

> Thanks merlimat , > i disabled auto-recovery component for bookies by running ookkeeper shell autorecovery -disable > and the issue is still happening looks like auto-recovery still trying to...