cmccabe commented on code in PR #17502:
URL: https://github.com/apache/kafka/pull/17502#discussion_r1811523833
##########
metadata/src/main/java/org/apache/kafka/controller/QuorumController.java:
##########
@@ -1855,6 +1633,97 @@ private QuorumController(
this.raftClient.register(metaLogListener);
}
+ /**
+ * Register the writeNoOpRecord task.
+ *
+ * This task periodically writes a NoOpRecord to the metadata log, if the
MetadataVersion
+ * supports it.
+ *
+ * @param maxIdleIntervalNs The period at which to write the
NoOpRecord.
+ */
+ private void registerWriteNoOpRecord(long maxIdleIntervalNs) {
+ periodicControl.registerTask(new PeriodicTask("writeNoOpRecord",
+ () -> {
+ ArrayList<ApiMessageAndVersion> records = new ArrayList<>(1);
+ if (featureControl.metadataVersion().isNoOpRecordSupported()) {
+ records.add(new ApiMessageAndVersion(new NoOpRecord(),
(short) 0));
+ }
+ return ControllerResult.of(records, false);
+ },
+ maxIdleIntervalNs,
+ EnumSet.noneOf(PeriodicTaskFlag.class)));
+ }
+
+ /**
+ * Calculate what the period should be for the maybeFenceStaleBroker task.
+ *
+ * @param sessionTimeoutNs The configured broker session timeout
period in nanoseconds.
+ *
+ * @return The period for the maybeFenceStaleBroker
task in nanoseconds.
+ */
+ static long maybeFenceStaleBrokerPeriodNs(long sessionTimeoutNs) {
+ return Math.max(TimeUnit.MILLISECONDS.toNanos(1), sessionTimeoutNs /
4);
Review Comment:
The more often we check, the more accurate our timeouts will be. The less
often we check, the more CPU we save. Using a fourth of the sessionTimeout
means that we'll have no more than 25% error. In practice this is checking
every 9.0 / 4 = 2.25 seconds which is not often. I suppose we could bump to 1/8
(every 1.125 seconds) to get error down to 12.5%...
--
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.
To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
