Hi Dima, They'll stick around. I think the point is that when node 3 is back, you should do partition reassignment so that the pre-existed partition/replica data can be reused.
But for your question, the unused log won't delete is a bug or as designed... I think it is working as designed and it should be removed manually. But that's just my thought. Thanks. Luke On Wed, Sep 16, 2020 at 10:48 AM Dima Brodsky <dbrod...@salesforce.com.invalid> wrote: > Hi, > > I have a question, when you start kafka on a node, if there is a random > replica log should it delete it on startup? Here is an example: Assume > you have a 4 node cluster. Topic X has 3 replicas and it is replicated on > nodes 1, 2, and 3. Now you shutdown node 3 and you place the replica that > was on node 3 on node 4. Then once everything is in-sync you start up node > 3 again. What should happen to the replica X on node 3. Should kafka > delete it or will it stick around forever. > > Given the above scenario we are seeing the replica stick around forever. > Is this working as designed, or is this a bug? > > Thanks! > ttyl > Dima > > -- > dbrod...@salesforce.com > > "The price of reliability is the pursuit of the utmost simplicity. > It is the price which the very rich find most hard to pay." (Sir Antony > Hoare, 1980) >
