public int completed(R request) {
assert request.equals(requests.peek());
requests.poll();
int remaining = requests.size();
if (remaining != 0) coordinator.send(requests.peek());
return remaining;
}
// we don't care about the return value but care about it throwing exception
public void runMayThrow() throws Exception {
if (endpoints.isEmpty()) {
differencingDone.signalAll();
logger.info(
"No neighbors to repair with for "
+ tablename
+ " on "
+ range
+ ": "
+ getName()
+ " completed.");
return;
}
// Checking all nodes are live
for (InetAddress endpoint : endpoints) {
if (!FailureDetector.instance.isAlive(endpoint)) {
differencingDone.signalAll();
logger.info(
"Could not proceed on repair because a neighbor ("
+ endpoint
+ ") is dead: "
+ getName()
+ " failed.");
return;
}
}
AntiEntropyService.instance.sessions.put(getName(), this);
Gossiper.instance.register(this);
FailureDetector.instance.registerFailureDetectionEventListener(this);
try {
// Create and queue a RepairJob for each column family
for (String cfname : cfnames) {
RepairJob job = new RepairJob(cfname);
jobs.offer(job);
activeJobs.put(cfname, job);
}
jobs.peek().sendTreeRequests();
// block whatever thread started this session until all requests have been returned:
// if this thread dies, the session will still complete in the background
completed.await();
if (exception != null) throw exception;
} catch (InterruptedException e) {
throw new RuntimeException(
"Interrupted while waiting for repair: repair will continue in the background.");
} finally {
FailureDetector.instance.unregisterFailureDetectionEventListener(this);
Gossiper.instance.unregister(this);
AntiEntropyService.instance.sessions.remove(getName());
}
}
void failedNode(InetAddress remote) {
String errorMsg =
String.format("Problem during repair session %s, endpoint %s died", sessionName, remote);
logger.error(errorMsg);
exception = new IOException(errorMsg);
// If a node failed, we stop everything (though there could still be some activity in the
// background)
jobs.clear();
activeJobs.clear();
differencingDone.signalAll();
completed.signalAll();
}
public void terminate() {
terminated = true;
for (RepairJob job : jobs) job.terminate();
jobs.clear();
activeJobs.clear();
}
// we don't care about the return value but care about it throwing exception
public void runMayThrow() throws Exception {
logger.info(
String.format(
"[repair #%s] new session: will sync %s on range %s for %s.%s",
getName(), repairedNodes(), range, tablename, Arrays.toString(cfnames)));
if (endpoints.isEmpty()) {
differencingDone.signalAll();
logger.info(
String.format(
"[repair #%s] No neighbors to repair with on range %s: session completed",
getName(), range));
return;
}
// Checking all nodes are live
for (InetAddress endpoint : endpoints) {
if (!FailureDetector.instance.isAlive(endpoint)) {
differencingDone.signalAll();
logger.info(
String.format(
"[repair #%s] Cannot proceed on repair because a neighbor (%s) is dead: session failed",
getName(), endpoint));
return;
}
if (Gossiper.instance.getVersion(endpoint) < MessagingService.VERSION_11 && isSequential) {
logger.info(
String.format(
"[repair #%s] Cannot repair using snapshots as node %s is pre-1.1",
getName(), endpoint));
return;
}
}
AntiEntropyService.instance.sessions.put(getName(), this);
Gossiper.instance.register(this);
FailureDetector.instance.registerFailureDetectionEventListener(this);
try {
// Create and queue a RepairJob for each column family
for (String cfname : cfnames) {
RepairJob job = new RepairJob(cfname);
jobs.offer(job);
activeJobs.put(cfname, job);
}
jobs.peek().sendTreeRequests();
// block whatever thread started this session until all requests have been returned:
// if this thread dies, the session will still complete in the background
completed.await();
if (exception == null) {
logger.info(String.format("[repair #%s] session completed successfully", getName()));
} else {
logger.error(
String.format("[repair #%s] session completed with the following error", getName()),
exception);
throw exception;
}
} catch (InterruptedException e) {
throw new RuntimeException("Interrupted while waiting for repair.");
} finally {
// mark this session as terminated
terminate();
FailureDetector.instance.unregisterFailureDetectionEventListener(this);
Gossiper.instance.unregister(this);
AntiEntropyService.instance.sessions.remove(getName());
}
}
public void start() {
if (requests.isEmpty()) return;
coordinator.send(requests.peek());
}
public void add(R request) {
requests.add(request);
}