[HDFS-16139] Update BPServiceActor Scheduler's nextBlockReportTime atomically

• Type: Task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.4.0, 3.3.5
• Fix Version/s: 3.4.0, 3.3.5
• Component/s: datanode
• Labels: pull-request-available
• Target Version/s: 3.4.0, 3.2.3, 3.3.2
• Hadoop Flags: Reviewed
• Assignee: Viraj Jasani
• Reporter: Viraj Jasani
• Created: 24/Jul/21 16:20
• Updated: 27/Jan/24 02:41
• Resolved: 27/Jul/21 06:59

Description

BPServiceActor#Scheduler's nextBlockReportTime is declared volatile and it can be assigned/read by testing threads (through BPServiceActor#triggerXXX) as well as by actor threads. Hence it is declared volatile but it is still assigned non-atomically

e.g

if (factor != 0) {
  nextBlockReportTime += factor * blockReportIntervalMs;
} else {
  // If the difference between the present time and the scheduled
  // time is very less, the factor can be 0, so in that case, we can
  // ignore that negligible time, spent while sending the BRss and
  // schedule the next BR after the blockReportInterval.
  nextBlockReportTime += blockReportIntervalMs;
}

We should convert it to AtomicLong to take care of concurrent assignments while making sure that it is assigned atomically.

Issue Links

• links to: GitHub Pull Request #3228