Conversation
|
@guozhangwang this should take care of your concern on KAFKA-2734. I like to keep separate concerns in separate commits, as it makes it easier to cherry-pick and follow. I hope it makes sense. Also, note that this branch already has KAFKA-2734, which is a pre-req to this PR, cherry-picked to help reviewers try out the changes locally. |
|
@SinghAsDev Could you rebase? |
ed75c47 to
b0d9ed5
Compare
|
@guozhangwang done |
There was a problem hiding this comment.
Could this happen with the check in ConsoleConsumer?
There was a problem hiding this comment.
This check is actually redundant, as it is also happening in ConsoleConsumer. However, the check just seemed to be natural here. Let me know if you are in favor of removing it.
There was a problem hiding this comment.
I was originally thinking if it should not ever happen let's remove it, after another thought I feel it may be OK to leave it as is.
|
LGTM, thanks. |
…ng loop (apache#445) This patch does the following: 1) Events are handled in order. The previous logic attempted to route the out-of-band registration/fencing messages to the front of the queue, but it seems simpler to let them be handled in order. I am not sure it is worthwhile optimizing for the case when there is a backlog of events to be processed. 2) Event failures are propagated. Previously we incremented an error counter, but an event failure leaves the broker state machine in an indeterminate state and it seems dangerous to try and continue. 3) Simplify processor API. Rather than overloading `process`, we can use a single method and let the processor `match` on the events it wants to handle. 4) Hook up `MetaLogManager.Listener`. The only real change here is that I preferred to retain the java List type since it awkwardly converts to the `Buffer` type with the scala collection converters.
…plicas on parition assignment cancellation (apache#445) This should be a direct fix-up to the commit - 28cca3e Cancellation of reassignments allowed if the resulting ISR is at min-ISR (apache#442)
…e#445) * feat(inkless): support loading google access token credentials If service account impersonation is used, then a plain access token is used for authentication. * fix(inkless): fix reloading Gcs credentials GoogleCredentials.equals returned true by comparing only public fields. Simply always reset Storage instance with given credentials.
…e#445) * feat(inkless): support loading google access token credentials If service account impersonation is used, then a plain access token is used for authentication. * fix(inkless): fix reloading Gcs credentials GoogleCredentials.equals returned true by comparing only public fields. Simply always reset Storage instance with given credentials.
#412 is a pre-req.