rhauch commented on a change in pull request #10822: URL: https://github.com/apache/kafka/pull/10822#discussion_r655683692
########## File path: connect/runtime/src/main/java/org/apache/kafka/connect/runtime/distributed/DistributedHerder.java ########## @@ -1592,6 +1731,28 @@ public void onSessionKeyUpdate(SessionKey sessionKey) { } } } + + @Override + public void onRestartRequest(RestartRequest request) { + log.info("Received and enqueuing {}", request); + + synchronized (DistributedHerder.this) { + String connectorName = request.connectorName(); + //preserve the highest impact request + if (pendingRestartRequests.containsKey(connectorName)) { + RestartRequest existingRequest = pendingRestartRequests.get(connectorName); + if (request.compareTo(existingRequest) > 0) { + log.debug("Overwriting existing {} and enqueuing the higher impact {}", existingRequest, request); + pendingRestartRequests.put(connectorName, request); + } else { + log.debug("Preserving existing higher impact {} and ignoring incoming {}", existingRequest, request); + } + } else { + pendingRestartRequests.put(connectorName, request); + } Review comment: If we were to change `pendingRestartRequests` to a `ConcurrentMap`, then we could use the `compute(...)` functionality that actually would work even if we removed the synchronization. I wonder if this is a bit more readable (even though strictly speaking we don't need concurrent access). WDYT? ``` pendingRestartRequests.compute(connectorName, (k, existing) -> { if (existing == null || request.compareTo(existing) > 0) { log.debug("Overwriting existing {} and enqueuing the higher impact {}", existing, request); return request; } log.debug("Preserving existing higher impact {} and ignoring incoming {}", existing, request); return existing; }); ``` -- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org