Search before asking

• I searched in the issues and found nothing similar.

Fluss version

0.6.0 (latest release)

Please describe the bug 🐞

When the tablet server fail, we should remove it from the ISR for the bucket that holds any replica in this server. It enables isr quick adjust instead of waiting util the follower lag for a period of time. Also will fix #644

Another problem will be:

1. server 0 is leader, server 1 is follower in isr
2. server 1 crash directly with page cache data loss.
3. server 0 controlled shut down.
4. server 1 start, server 1 will be elected as leader. So the data in page cache of server1 of step 1 will lose

The data is loss..

Solution

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!