Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

ZOOKEEPER-2678: Discovery and Sync can take a very long time on large DB (3.4) #157

Closed
wants to merge 5 commits into from
Closed

ZOOKEEPER-2678: Discovery and Sync can take a very long time on large DB (3.4) #157

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
5aa2562
ZOOKEEPER-2678: Discovery and Sync can take a very long time on large…
Jan 19, 2017
f705293
ZOOKEEPER-2678: Addressed review comments
Jan 30, 2017
f57c384
ZOOKEEPER-2678: Fixed another race
Jan 30, 2017
dcbf325
Addressed review comments
Jan 31, 2017
d079617
ZOOKEEPER-2678: Improved test to verify snapshot times
Feb 6, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 16 additions & 2 deletions src/java/main/org/apache/zookeeper/server/ZooKeeperServer.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -489,7 +489,15 @@ public boolean isRunning() {
return state == State.RUNNING;
}

public synchronized void shutdown() {
public void shutdown() {
shutdown(false);
}

/**
* Shut down the server instance
* @param fullyShutDown true if another server using the same database will not replace this one in the same process
*/
public synchronized void shutdown(boolean fullyShutDown) {
if (!canShutdown()) {
LOG.debug("ZooKeeper server is not running, so not proceeding to shutdown!");
return;
Expand All @@ -507,9 +515,15 @@ public synchronized void shutdown() {
if (firstProcessor != null) {
firstProcessor.shutdown();
}
if (zkDb != null) {

if (fullyShutDown && zkDb != null) {
zkDb.clear();
}
// else there is no need to clear the database
// * When a new quorum is established we can still apply the diff
// on top of the same zkDb data
// * If we fetch a new snapshot from leader, the zkDb will be
// cleared anyway before loading the snapshot

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There is one case we may still want to clear db here - when one of the ZooKeeper critical threads (such as * processors, session trackers) fail, ZooKeeper server will shutdown (see runFromConfig) and consequently invoke ZooKeeper#shutdown. In this case, I don't see a particular reason not to clear the db, though not doing it might be fine (as one could argue the server will be dead anyway), but I tend to lean towards the safe side on cleaning the db. One way to conditionally do that is to add a Boolean parameter to ZooKeeper#shutdown so we can have fine grained control over when to clear db in what code path.

unregisterJMX();
}
Expand Down
2 changes: 1 addition & 1 deletion src/java/main/org/apache/zookeeper/server/ZooKeeperServerMain.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -124,7 +124,7 @@ public void runFromConfig(ServerConfig config) throws IOException {

cnxnFactory.join();
if (zkServer.canShutdown()) {
zkServer.shutdown();
zkServer.shutdown(true);
}
} catch (InterruptedException e) {
// warn, but generally this is ok
Expand Down
34 changes: 22 additions & 12 deletions src/java/main/org/apache/zookeeper/server/quorum/Learner.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -321,13 +321,16 @@ protected void syncWithLeader(long newLeaderZxid) throws IOException, Interrupte
QuorumPacket ack = new QuorumPacket(Leader.ACK, 0, null, null);
QuorumPacket qp = new QuorumPacket();
long newEpoch = ZxidUtils.getEpochFromZxid(newLeaderZxid);

readPacket(qp);
// In the DIFF case we don't need to do a snapshot because the transactions will sync on top of any existing snapshot
// For SNAP and TRUNC the snapshot is needed to save that history
boolean snapshotNeeded = true;
readPacket(qp);
LinkedList<Long> packetsCommitted = new LinkedList<Long>();
LinkedList<PacketInFlight> packetsNotCommitted = new LinkedList<PacketInFlight>();
synchronized (zk) {
if (qp.getType() == Leader.DIFF) {
LOG.info("Getting a diff from the leader 0x" + Long.toHexString(qp.getZxid()));
LOG.info("Getting a diff from the leader 0x{}", Long.toHexString(qp.getZxid()));
snapshotNeeded = false;
}
else if (qp.getType() == Leader.SNAP) {
LOG.info("Getting a snapshot from leader");
Expand Down Expand Up @@ -364,10 +367,13 @@ else if (qp.getType() == Leader.SNAP) {

long lastQueued = 0;

// in V1.0 we take a snapshot when we get the NEWLEADER message, but in pre V1.0
// we take the snapshot at the UPDATE, since V1.0 also gets the UPDATE (after the NEWLEADER)
// in Zab V1.0 (ZK 3.4+) we might take a snapshot when we get the NEWLEADER message, but in pre V1.0
// we take the snapshot on the UPDATE message, since Zab V1.0 also gets the UPDATE (after the NEWLEADER)
// we need to make sure that we don't take the snapshot twice.
boolean snapshotTaken = false;
boolean isPreZAB1_0 = true;
//If we are not going to take the snapshot be sure the transactions are not applied in memory
// but written out to the transaction log
boolean writeToTxnLog = !snapshotNeeded;
// we are now going to start getting transactions to apply followed by an UPTODATE
outerLoop:
while (self.isRunning()) {
Expand All @@ -387,7 +393,7 @@ else if (qp.getType() == Leader.SNAP) {
packetsNotCommitted.add(pif);
break;
case Leader.COMMIT:
if (!snapshotTaken) {
if (!writeToTxnLog) {
pif = packetsNotCommitted.peekFirst();
if (pif.hdr.getZxid() != qp.getZxid()) {
LOG.warn("Committing " + qp.getZxid() + ", but next proposal is " + pif.hdr.getZxid());
Expand Down Expand Up @@ -415,7 +421,7 @@ else if (qp.getType() == Leader.SNAP) {
+ Long.toHexString(lastQueued + 1));
}
lastQueued = packet.hdr.getZxid();
if (!snapshotTaken) {
if (!writeToTxnLog) {
// Apply to db directly if we haven't taken the snapshot
zk.processTxn(packet.hdr, packet.rec);
} else {
Expand All @@ -424,13 +430,14 @@ else if (qp.getType() == Leader.SNAP) {
}
break;
case Leader.UPTODATE:
if (!snapshotTaken) { // true for the pre v1.0 case
if (isPreZAB1_0) {
zk.takeSnapshot();
self.setCurrentEpoch(newEpoch);
}
self.cnxnFactory.setZooKeeperServer(zk);
break outerLoop;
case Leader.NEWLEADER: // it will be NEWLEADER in v1.0
case Leader.NEWLEADER: // Getting NEWLEADER here instead of in discovery
// means this is Zab 1.0
// Create updatingEpoch file and remove it after current
// epoch is set. QuorumPeer.loadDataBase() uses this file to
// detect the case where the server was terminated after
Expand All @@ -441,13 +448,16 @@ else if (qp.getType() == Leader.SNAP) {
throw new IOException("Failed to create " +
updating.toString());
}
zk.takeSnapshot();
if (snapshotNeeded) {
zk.takeSnapshot();
}
self.setCurrentEpoch(newEpoch);
if (!updating.delete()) {
throw new IOException("Failed to delete " +
updating.toString());
}
snapshotTaken = true;
writeToTxnLog = true; //Anything after this needs to go to the transaction log, not applied directly in memory
isPreZAB1_0 = false;
writePacket(new QuorumPacket(Leader.ACK, newLeaderZxid, null, null), true);
break;
}
Expand Down
29 changes: 28 additions & 1 deletion src/java/test/org/apache/zookeeper/server/quorum/Zab1_0Test.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,9 @@
package org.apache.zookeeper.server.quorum;

import static org.junit.Assert.assertEquals;
import static org.mockito.Mockito.never;
import static org.mockito.Mockito.spy;
import static org.mockito.Mockito.verify;

import java.io.BufferedInputStream;
import java.io.BufferedReader;
Expand Down Expand Up @@ -645,6 +648,8 @@ public void converseWithFollower(InputArchive ia, OutputArchive oa,
tmpDir.mkdir();
File logDir = f.fzk.getTxnLogFactory().getDataDir().getParentFile();
File snapDir = f.fzk.getTxnLogFactory().getSnapDir().getParentFile();
//Spy on ZK so we can check if a snapshot happened or not.
f.zk = spy(f.zk);
try {
Assert.assertEquals(0, f.self.getAcceptedEpoch());
Assert.assertEquals(0, f.self.getCurrentEpoch());
Expand Down Expand Up @@ -687,6 +692,10 @@ public void converseWithFollower(InputArchive ia, OutputArchive oa,
oa.writeRecord(qp, null);
zkDb.serializeSnapshot(oa);
oa.writeString("BenWasHere", null);
Thread.sleep(10); //Give it some time to process the snap
//No Snapshot taken yet, the SNAP was applied in memory
verify(f.zk, never()).takeSnapshot();

qp.setType(Leader.NEWLEADER);
qp.setZxid(ZxidUtils.makeZxid(1, 0));
oa.writeRecord(qp, null);
Expand All @@ -697,7 +706,8 @@ public void converseWithFollower(InputArchive ia, OutputArchive oa,
Assert.assertEquals(ZxidUtils.makeZxid(1, 0), qp.getZxid());
Assert.assertEquals(1, f.self.getAcceptedEpoch());
Assert.assertEquals(1, f.self.getCurrentEpoch());

//Make sure that we did take the snapshot now
verify(f.zk).takeSnapshot();
Assert.assertEquals(firstZxid, f.fzk.getLastProcessedZxid());

// Make sure the data was recorded in the filesystem ok
Expand Down Expand Up @@ -773,6 +783,8 @@ public void converseWithFollower(InputArchive ia, OutputArchive oa,
tmpDir.mkdir();
File logDir = f.fzk.getTxnLogFactory().getDataDir().getParentFile();
File snapDir = f.fzk.getTxnLogFactory().getSnapDir().getParentFile();
//Spy on ZK so we can check if a snapshot happened or not.
f.zk = spy(f.zk);
try {
Assert.assertEquals(0, f.self.getAcceptedEpoch());
Assert.assertEquals(0, f.self.getCurrentEpoch());
Expand Down Expand Up @@ -839,13 +851,28 @@ public void converseWithFollower(InputArchive ia, OutputArchive oa,
Assert.assertEquals(1, f.self.getAcceptedEpoch());
Assert.assertEquals(1, f.self.getCurrentEpoch());

//Wait for the transactions to be written out. The thread that writes them out
// does not send anything back when it is done.
long start = System.currentTimeMillis();
while (createSessionZxid != f.fzk.getLastProcessedZxid() && (System.currentTimeMillis() - start) < 50) {
Copy link
Contributor

@afine afine Jan 31, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

just an idea, not sure if it is worth the effort and it may be outside the scope of this patch.

we could play with the test infrastructure here a little bit and do some dependency injection in createFollower that can let us track if db clearing and snapshotting occurs when expected.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It does seem a bit beyond the scope of this. But if you really want me to I can look into it.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

it would be nice, a good way of actually validating everything is behaving as expected

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

OK Will see what I can do

Thread.sleep(1);
}

Assert.assertEquals(createSessionZxid, f.fzk.getLastProcessedZxid());

// Make sure the data was recorded in the filesystem ok
ZKDatabase zkDb2 = new ZKDatabase(new FileTxnSnapLog(logDir, snapDir));
start = System.currentTimeMillis();
zkDb2.loadDataBase();
while (zkDb2.getSessionWithTimeOuts().isEmpty() && (System.currentTimeMillis() - start) < 50) {
Thread.sleep(1);
zkDb2.loadDataBase();
}
LOG.info("zkdb2 sessions:" + zkDb2.getSessions());
LOG.info("zkdb2 with timeouts:" + zkDb2.getSessionWithTimeOuts());
Assert.assertNotNull(zkDb2.getSessionWithTimeOuts().get(4L));
//Snapshot was never taken during very simple sync
verify(f.zk, never()).takeSnapshot();
} finally {
recursiveDelete(tmpDir);
}
Expand Down