etan39
/
cloudstack
mirror of https://github.com/apache/cloudstack.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
cloudstack/framework/cluster/src/com/cloud/cluster/ClusterManagerImpl.java

1188 lines
48 KiB
Java
Raw Blame History

// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing,
// software distributed under the License is distributed on an
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied. See the License for the
// specific language governing permissions and limitations
// under the License.
package com.cloud.cluster;
import java.io.IOException;
import java.net.ConnectException;
import java.net.InetSocketAddress;
import java.nio.channels.SocketChannel;
import java.rmi.RemoteException;
import java.sql.Connection;
import java.sql.SQLException;
import java.sql.SQLNonTransientException;
import java.sql.SQLRecoverableException;
import java.util.ArrayList;
import java.util.Date;
import java.util.HashMap;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Properties;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.TimeUnit;
import javax.ejb.Local;
import javax.inject.Inject;
import javax.naming.ConfigurationException;
import org.apache.cloudstack.framework.config.ConfigDepot;
import org.apache.cloudstack.framework.config.ConfigKey;
import org.apache.cloudstack.framework.config.Configurable;
import org.apache.cloudstack.managed.context.ManagedContextRunnable;
import org.apache.cloudstack.utils.identity.ManagementServerNode;
import org.apache.log4j.Logger;
import com.cloud.cluster.dao.ManagementServerHostDao;
import com.cloud.cluster.dao.ManagementServerHostPeerDao;
import com.cloud.utils.DateUtil;
import com.cloud.utils.Profiler;
import com.cloud.utils.component.ComponentLifecycle;
import com.cloud.utils.component.ManagerBase;
import com.cloud.utils.concurrency.NamedThreadFactory;
import com.cloud.utils.db.ConnectionConcierge;
import com.cloud.utils.db.DB;
import com.cloud.utils.db.DbProperties;
import com.cloud.utils.db.Transaction;
import com.cloud.utils.db.TransactionCallback;
import com.cloud.utils.db.TransactionLegacy;
import com.cloud.utils.db.TransactionStatus;
import com.cloud.utils.events.SubscriptionMgr;
import com.cloud.utils.exception.CloudRuntimeException;
import com.cloud.utils.exception.ExceptionUtil;
import com.cloud.utils.mgmt.JmxUtil;
import com.cloud.utils.net.NetUtils;
@Local(value = {ClusterManager.class})
public class ClusterManagerImpl extends ManagerBase implements ClusterManager, Configurable {
private static final Logger s_logger = Logger.getLogger(ClusterManagerImpl.class);
private static final int EXECUTOR_SHUTDOWN_TIMEOUT = 1000; // 1 second
private static final int DEFAULT_OUTGOING_WORKERS = 5;
private final List<ClusterManagerListener> _listeners = new ArrayList<ClusterManagerListener>();
private final Map<Long, ManagementServerHostVO> _activePeers = new HashMap<Long, ManagementServerHostVO>();
private final Map<String, ClusterService> _clusterPeers;
@Inject
protected ConfigDepot _configDepot;
private final ScheduledExecutorService _heartbeatScheduler = Executors.newScheduledThreadPool(1, new NamedThreadFactory("Cluster-Heartbeat"));
private final ExecutorService _notificationExecutor = Executors.newFixedThreadPool(1, new NamedThreadFactory("Cluster-Notification"));
private final List<ClusterManagerMessage> _notificationMsgs = new ArrayList<ClusterManagerMessage>();
private ConnectionConcierge _heartbeatConnection = null;
private final ExecutorService _executor;
private ClusterServiceAdapter _currentServiceAdapter;
@Inject
private List<ClusterServiceAdapter> _serviceAdapters;
@Inject
private ManagementServerHostDao _mshostDao;
@Inject
private ManagementServerHostPeerDao _mshostPeerDao;
protected Dispatcher _dispatcher;
//
// pay attention to _mshostId and _msid
// _mshostId is the primary key of management host table
// _msid is the unique persistent identifier that peer name is based upon
//
private Long _mshostId = null;
protected long _msId = ManagementServerNode.getManagementServerId();
protected long _runId = System.currentTimeMillis();
private boolean _peerScanInited = false;
private String _clusterNodeIP = "127.0.0.1";
private final List<ClusterServicePdu> _clusterPduOutgoingQueue = new ArrayList<ClusterServicePdu>();
private final List<ClusterServicePdu> _clusterPduIncomingQueue = new ArrayList<ClusterServicePdu>();
private final Map<Long, ClusterServiceRequestPdu> _outgoingPdusWaitingForAck = new HashMap<Long, ClusterServiceRequestPdu>();
public ClusterManagerImpl() {
_clusterPeers = new HashMap<String, ClusterService>();
// executor to perform remote-calls in another thread context, to avoid potential
// recursive remote calls between nodes
//
_executor = Executors.newCachedThreadPool(new NamedThreadFactory("Cluster-Worker"));
setRunLevel(ComponentLifecycle.RUN_LEVEL_FRAMEWORK);
}
private void registerRequestPdu(ClusterServiceRequestPdu pdu) {
synchronized (_outgoingPdusWaitingForAck) {
_outgoingPdusWaitingForAck.put(pdu.getSequenceId(), pdu);
}
}
@Override
public void registerDispatcher(Dispatcher dispatcher) {
_dispatcher = dispatcher;
}
private ClusterServiceRequestPdu popRequestPdu(long ackSequenceId) {
synchronized (_outgoingPdusWaitingForAck) {
if (_outgoingPdusWaitingForAck.get(ackSequenceId) != null) {
ClusterServiceRequestPdu pdu = _outgoingPdusWaitingForAck.get(ackSequenceId);
_outgoingPdusWaitingForAck.remove(ackSequenceId);
return pdu;
}
}
return null;
}
private void cancelClusterRequestToPeer(String strPeer) {
List<ClusterServiceRequestPdu> candidates = new ArrayList<ClusterServiceRequestPdu>();
synchronized (_outgoingPdusWaitingForAck) {
for (Map.Entry<Long, ClusterServiceRequestPdu> entry : _outgoingPdusWaitingForAck.entrySet()) {
if (entry.getValue().getDestPeer().equalsIgnoreCase(strPeer))
candidates.add(entry.getValue());
}
for (ClusterServiceRequestPdu pdu : candidates) {
_outgoingPdusWaitingForAck.remove(pdu.getSequenceId());
}
}
for (ClusterServiceRequestPdu pdu : candidates) {
s_logger.warn("Cancel cluster request PDU to peer: " + strPeer + ", pdu: " + pdu.getJsonPackage());
synchronized (pdu) {
pdu.notifyAll();
}
}
}
private void addOutgoingClusterPdu(ClusterServicePdu pdu) {
synchronized (_clusterPduOutgoingQueue) {
_clusterPduOutgoingQueue.add(pdu);
_clusterPduOutgoingQueue.notifyAll();
}
}
private ClusterServicePdu popOutgoingClusterPdu(long timeoutMs) {
synchronized (_clusterPduOutgoingQueue) {
try {
_clusterPduOutgoingQueue.wait(timeoutMs);
} catch (InterruptedException e) {
}
if (_clusterPduOutgoingQueue.size() > 0) {
ClusterServicePdu pdu = _clusterPduOutgoingQueue.get(0);
_clusterPduOutgoingQueue.remove(0);
return pdu;
}
}
return null;
}
private void addIncomingClusterPdu(ClusterServicePdu pdu) {
synchronized (_clusterPduIncomingQueue) {
_clusterPduIncomingQueue.add(pdu);
_clusterPduIncomingQueue.notifyAll();
}
}
private ClusterServicePdu popIncomingClusterPdu(long timeoutMs) {
synchronized (_clusterPduIncomingQueue) {
try {
_clusterPduIncomingQueue.wait(timeoutMs);
} catch (InterruptedException e) {
}
if (_clusterPduIncomingQueue.size() > 0) {
ClusterServicePdu pdu = _clusterPduIncomingQueue.get(0);
_clusterPduIncomingQueue.remove(0);
return pdu;
}
}
return null;
}
private Runnable getClusterPduSendingTask() {
return new ManagedContextRunnable() {
@Override
protected void runInContext() {
onSendingClusterPdu();
}
};
}
private Runnable getClusterPduNotificationTask() {
return new ManagedContextRunnable() {
@Override
protected void runInContext() {
onNotifyingClusterPdu();
}
};
}
private void onSendingClusterPdu() {
while (true) {
try {
ClusterServicePdu pdu = popOutgoingClusterPdu(1000);
if (pdu == null)
continue;
ClusterService peerService = null;
for (int i = 0; i < 2; i++) {
try {
peerService = getPeerService(pdu.getDestPeer());
} catch (RemoteException e) {
s_logger.error("Unable to get cluster service on peer : " + pdu.getDestPeer());
}
if (peerService != null) {
try {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Cluster PDU " + getSelfPeerName() + " -> " + pdu.getDestPeer() + ". agent: " + pdu.getAgentId() + ", pdu seq: " +
pdu.getSequenceId() + ", pdu ack seq: " + pdu.getAckSequenceId() + ", json: " + pdu.getJsonPackage());
}
long startTick = System.currentTimeMillis();
String strResult = peerService.execute(pdu);
if (s_logger.isDebugEnabled()) {
s_logger.debug("Cluster PDU " + getSelfPeerName() + " -> " + pdu.getDestPeer() + " completed. time: " +
(System.currentTimeMillis() - startTick) + "ms. agent: " + pdu.getAgentId() + ", pdu seq: " + pdu.getSequenceId() +
", pdu ack seq: " + pdu.getAckSequenceId() + ", json: " + pdu.getJsonPackage());
}
if ("true".equals(strResult))
break;
} catch (RemoteException e) {
invalidatePeerService(pdu.getDestPeer());
if (s_logger.isInfoEnabled()) {
s_logger.info("Exception on remote execution, peer: " + pdu.getDestPeer() + ", iteration: " + i + ", exception message :" +
e.getMessage());
}
}
}
}
} catch (Throwable e) {
s_logger.error("Unexcpeted exception: ", e);
}
}
}
private void onNotifyingClusterPdu() {
while (true) {
try {
final ClusterServicePdu pdu = popIncomingClusterPdu(1000);
if (pdu == null)
continue;
_executor.execute(new ManagedContextRunnable() {
@Override
protected void runInContext() {
if (pdu.getPduType() == ClusterServicePdu.PDU_TYPE_RESPONSE) {
ClusterServiceRequestPdu requestPdu = popRequestPdu(pdu.getAckSequenceId());
if (requestPdu != null) {
requestPdu.setResponseResult(pdu.getJsonPackage());
synchronized (requestPdu) {
requestPdu.notifyAll();
}
} else {
s_logger.warn("Original request has already been cancelled. pdu: " + pdu.getJsonPackage());
}
} else {
String result = _dispatcher.dispatch(pdu);
if (result == null)
result = "";
if (pdu.getPduType() == ClusterServicePdu.PDU_TYPE_REQUEST) {
ClusterServicePdu responsePdu = new ClusterServicePdu();
responsePdu.setPduType(ClusterServicePdu.PDU_TYPE_RESPONSE);
responsePdu.setSourcePeer(pdu.getDestPeer());
responsePdu.setDestPeer(pdu.getSourcePeer());
responsePdu.setAckSequenceId(pdu.getSequenceId());
responsePdu.setJsonPackage(result);
addOutgoingClusterPdu(responsePdu);
}
}
}
});
} catch (Throwable e) {
s_logger.error("Unexcpeted exception: ", e);
}
}
}
@Override
public void OnReceiveClusterServicePdu(ClusterServicePdu pdu) {
addIncomingClusterPdu(pdu);
}
/**
* called by DatabaseUpgradeChecker to see if there are other peers running.
*
* @param notVersion
* If version is passed in, the peers CANNOT be running at this version. If version is null, return true if any
* peer is running regardless of version.
* @return true if there are peers running and false if not.
*/
public static final boolean arePeersRunning(String notVersion) {
return false; // TODO: Leaving this for Kelven to take care of.
}
@Override
public void broadcast(long agentId, String cmds) {
Date cutTime = DateUtil.currentGMTTime();
List<ManagementServerHostVO> peers = _mshostDao.getActiveList(new Date(cutTime.getTime() - HeartbeatThreshold.value()));
for (ManagementServerHostVO peer : peers) {
String peerName = Long.toString(peer.getMsid());
if (getSelfPeerName().equals(peerName)) {
continue; // Skip myself.
}
try {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Forwarding " + cmds + " to " + peer.getMsid());
}
executeAsync(peerName, agentId, cmds, true);
} catch (Exception e) {
s_logger.warn("Caught exception while talkign to " + peer.getMsid());
}
}
}
public void executeAsync(String strPeer, long agentId, String cmds, boolean stopOnError) {
ClusterServicePdu pdu = new ClusterServicePdu();
pdu.setSourcePeer(getSelfPeerName());
pdu.setDestPeer(strPeer);
pdu.setAgentId(agentId);
pdu.setJsonPackage(cmds);
pdu.setStopOnError(true);
addOutgoingClusterPdu(pdu);
}
@Override
public String execute(String strPeer, long agentId, String cmds, boolean stopOnError) {
if (s_logger.isDebugEnabled()) {
s_logger.debug(getSelfPeerName() + " -> " + strPeer + "." + agentId + " " + cmds);
}
ClusterServiceRequestPdu pdu = new ClusterServiceRequestPdu();
pdu.setSourcePeer(getSelfPeerName());
pdu.setDestPeer(strPeer);
pdu.setAgentId(agentId);
pdu.setJsonPackage(cmds);
pdu.setStopOnError(stopOnError);
registerRequestPdu(pdu);
addOutgoingClusterPdu(pdu);
synchronized (pdu) {
try {
pdu.wait();
} catch (InterruptedException e) {
}
}
if (s_logger.isDebugEnabled()) {
s_logger.debug(getSelfPeerName() + " -> " + strPeer + "." + agentId + " completed. result: " + pdu.getResponseResult());
}
if (pdu.getResponseResult() != null && pdu.getResponseResult().length() > 0) {
return pdu.getResponseResult();
}
return null;
}
@Override
public ManagementServerHostVO getPeer(String mgmtServerId) {
return _mshostDao.findByMsid(Long.valueOf(mgmtServerId));
}
@Override
public String getSelfPeerName() {
return Long.toString(_msId);
}
public String getSelfNodeIP() {
return _clusterNodeIP;
}
@Override
public void registerListener(ClusterManagerListener listener) {
// Note : we don't check duplicates
synchronized (_listeners) {
s_logger.info("register cluster listener " + listener.getClass());
_listeners.add(listener);
}
}
@Override
public void unregisterListener(ClusterManagerListener listener) {
synchronized (_listeners) {
s_logger.info("unregister cluster listener " + listener.getClass());
_listeners.remove(listener);
}
}
public void notifyNodeJoined(List<ManagementServerHostVO> nodeList) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Notify management server node join to listeners.");
for (ManagementServerHostVO mshost : nodeList) {
s_logger.debug("Joining node, IP: " + mshost.getServiceIP() + ", msid: " + mshost.getMsid());
}
}
synchronized (_listeners) {
for (ClusterManagerListener listener : _listeners) {
listener.onManagementNodeJoined(nodeList, _mshostId);
}
}
SubscriptionMgr.getInstance().notifySubscribers(ClusterManager.ALERT_SUBJECT, this, new ClusterNodeJoinEventArgs(_mshostId, nodeList));
}
public void notifyNodeLeft(List<ManagementServerHostVO> nodeList) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Notify management server node left to listeners.");
}
for (ManagementServerHostVO mshost : nodeList) {
if (s_logger.isDebugEnabled())
s_logger.debug("Leaving node, IP: " + mshost.getServiceIP() + ", msid: " + mshost.getMsid());
cancelClusterRequestToPeer(String.valueOf(mshost.getMsid()));
}
synchronized (_listeners) {
for (ClusterManagerListener listener : _listeners) {
listener.onManagementNodeLeft(nodeList, _mshostId);
}
}
SubscriptionMgr.getInstance().notifySubscribers(ClusterManager.ALERT_SUBJECT, this, new ClusterNodeLeftEventArgs(_mshostId, nodeList));
}
public void notifyNodeIsolated() {
if (s_logger.isDebugEnabled())
s_logger.debug("Notify management server node isolation to listeners");
synchronized (_listeners) {
for (ClusterManagerListener listener : _listeners) {
listener.onManagementNodeIsolated();
}
}
}
public ClusterService getPeerService(String strPeer) throws RemoteException {
synchronized (_clusterPeers) {
if (_clusterPeers.containsKey(strPeer)) {
return _clusterPeers.get(strPeer);
}
}
ClusterService service = _currentServiceAdapter.getPeerService(strPeer);
if (service != null) {
synchronized (_clusterPeers) {
// re-check the peer map again to deal with the
// race conditions
if (!_clusterPeers.containsKey(strPeer)) {
_clusterPeers.put(strPeer, service);
}
}
}
return service;
}
public void invalidatePeerService(String strPeer) {
synchronized (_clusterPeers) {
if (_clusterPeers.containsKey(strPeer)) {
_clusterPeers.remove(strPeer);
}
}
}
private Runnable getHeartbeatTask() {
return new ManagedContextRunnable() {
@Override
protected void runInContext() {
TransactionLegacy txn = TransactionLegacy.open("ClusterHeartbeat");
try {
Profiler profiler = new Profiler();
Profiler profilerHeartbeatUpdate = new Profiler();
Profiler profilerPeerScan = new Profiler();
try {
profiler.start();
profilerHeartbeatUpdate.start();
txn.transitToUserManagedConnection(getHeartbeatConnection());
if (s_logger.isTraceEnabled()) {
s_logger.trace("Cluster manager heartbeat update, id:" + _mshostId);
}
_mshostDao.update(_mshostId, getCurrentRunId(), DateUtil.currentGMTTime());
profilerHeartbeatUpdate.stop();
profilerPeerScan.start();
if (s_logger.isTraceEnabled()) {
s_logger.trace("Cluster manager peer-scan, id:" + _mshostId);
}
if (!_peerScanInited) {
_peerScanInited = true;
initPeerScan();
}
peerScan();
profilerPeerScan.stop();
} finally {
profiler.stop();
if (profiler.getDuration() >= HeartbeatInterval.value()) {
if (s_logger.isDebugEnabled())
s_logger.debug("Management server heartbeat takes too long to finish. profiler: " + profiler.toString() + ", profilerHeartbeatUpdate: " +
profilerHeartbeatUpdate.toString() + ", profilerPeerScan: " + profilerPeerScan.toString());
}
}
} catch (CloudRuntimeException e) {
s_logger.error("Runtime DB exception ", e.getCause());
if (e.getCause() instanceof ClusterInvalidSessionException) {
s_logger.error("Invalid cluster session found, fence it");
queueNotification(new ClusterManagerMessage(ClusterManagerMessage.MessageType.nodeIsolated));
}
if (isRootCauseConnectionRelated(e.getCause())) {
invalidHeartbeatConnection();
}
} catch (ActiveFencingException e) {
queueNotification(new ClusterManagerMessage(ClusterManagerMessage.MessageType.nodeIsolated));
} catch (Throwable e) {
s_logger.error("Unexpected exception in cluster heartbeat", e);
if (isRootCauseConnectionRelated(e.getCause())) {
invalidHeartbeatConnection();
}
} finally {
txn.transitToAutoManagedConnection(TransactionLegacy.CLOUD_DB);
txn.close("ClusterHeartbeat");
}
}
};
}
private boolean isRootCauseConnectionRelated(Throwable e) {
while (e != null) {
if (e instanceof SQLRecoverableException || e instanceof SQLNonTransientException) {
return true;
}
e = e.getCause();
}
return false;
}
private Connection getHeartbeatConnection() throws SQLException {
if (_heartbeatConnection == null) {
Connection conn = TransactionLegacy.getStandaloneConnectionWithException();
_heartbeatConnection = new ConnectionConcierge("ClusterManagerHeartbeat", conn, false);
}
return _heartbeatConnection.conn();
}
private void invalidHeartbeatConnection() {
if (_heartbeatConnection != null) {
Connection conn = TransactionLegacy.getStandaloneConnection();
if (conn != null) {
_heartbeatConnection.reset(conn);
} else {
s_logger.error("DB communication problem detected, fence it");
queueNotification(new ClusterManagerMessage(ClusterManagerMessage.MessageType.nodeIsolated));
}
// The stand-alone connection does not have to be closed here because there will be another reference to it.
// As a matter of fact, it will be assigned to the connection instance variable in the ConnectionConcierge class.
}
}
private Runnable getNotificationTask() {
return new ManagedContextRunnable() {
@Override
protected void runInContext() {
while (true) {
synchronized (_notificationMsgs) {
try {
_notificationMsgs.wait(1000);
} catch (InterruptedException e) {
}
}
ClusterManagerMessage msg = null;
while ((msg = getNextNotificationMessage()) != null) {
try {
switch (msg.getMessageType()) {
case nodeAdded:
if (msg.getNodes() != null && msg.getNodes().size() > 0) {
Profiler profiler = new Profiler();
profiler.start();
notifyNodeJoined(msg.getNodes());
profiler.stop();
if (profiler.getDuration() > 1000) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Notifying management server join event took " + profiler.getDuration() + " ms");
}
} else {
s_logger.warn("Notifying management server join event took " + profiler.getDuration() + " ms");
}
}
break;
case nodeRemoved:
if (msg.getNodes() != null && msg.getNodes().size() > 0) {
Profiler profiler = new Profiler();
profiler.start();
notifyNodeLeft(msg.getNodes());
profiler.stop();
if (profiler.getDuration() > 1000) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Notifying management server leave event took " + profiler.getDuration() + " ms");
}
} else {
s_logger.warn("Notifying management server leave event took " + profiler.getDuration() + " ms");
}
}
break;
case nodeIsolated:
notifyNodeIsolated();
break;
default:
assert (false);
break;
}
} catch (Throwable e) {
s_logger.warn("Unexpected exception during cluster notification. ", e);
}
}
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
}
}
}
};
}
private void queueNotification(ClusterManagerMessage msg) {
synchronized (_notificationMsgs) {
_notificationMsgs.add(msg);
_notificationMsgs.notifyAll();
}
switch (msg.getMessageType()) {
case nodeAdded: {
List<ManagementServerHostVO> l = msg.getNodes();
if (l != null && l.size() > 0) {
for (ManagementServerHostVO mshost : l) {
_mshostPeerDao.updatePeerInfo(_mshostId, mshost.getId(), mshost.getRunid(), ManagementServerHost.State.Up);
}
}
}
break;
case nodeRemoved: {
List<ManagementServerHostVO> l = msg.getNodes();
if (l != null && l.size() > 0) {
for (ManagementServerHostVO mshost : l) {
_mshostPeerDao.updatePeerInfo(_mshostId, mshost.getId(), mshost.getRunid(), ManagementServerHost.State.Down);
}
}
}
break;
default:
break;
}
}
private ClusterManagerMessage getNextNotificationMessage() {
synchronized (_notificationMsgs) {
if (_notificationMsgs.size() > 0) {
return _notificationMsgs.remove(0);
}
}
return null;
}
private void initPeerScan() {
// upon startup, for all inactive management server nodes that we see at startup time, we will send notification also to help upper layer perform
// missed cleanup
Date cutTime = DateUtil.currentGMTTime();
List<ManagementServerHostVO> inactiveList = _mshostDao.getInactiveList(new Date(cutTime.getTime() - HeartbeatThreshold.value()));
// We don't have foreign key constraints to enforce the mgmt_server_id integrity in host table, when user manually
// remove records from mshost table, this will leave orphan mgmt_serve_id reference in host table.
List<Long> orphanList = _mshostDao.listOrphanMsids();
if (orphanList.size() > 0) {
for (Long orphanMsid : orphanList) {
// construct fake ManagementServerHostVO based on orphan MSID
s_logger.info("Add orphan management server msid found in host table to initial clustering notification, orphan msid: " + orphanMsid);
inactiveList.add(new ManagementServerHostVO(orphanMsid, 0, "orphan", 0, new Date()));
}
} else {
s_logger.info("We are good, no orphan management server msid in host table is found");
}
if (inactiveList.size() > 0) {
if (s_logger.isInfoEnabled()) {
s_logger.info("Found " + inactiveList.size() + " inactive management server node based on timestamp");
for (ManagementServerHostVO host : inactiveList)
s_logger.info("management server node msid: " + host.getMsid() + ", name: " + host.getName() + ", service ip: " + host.getServiceIP() +
", version: " + host.getVersion());
}
List<ManagementServerHostVO> downHostList = new ArrayList<ManagementServerHostVO>();
for (ManagementServerHostVO host : inactiveList) {
if (!pingManagementNode(host)) {
s_logger.warn("Management node " + host.getId() + " is detected inactive by timestamp and also not pingable");
downHostList.add(host);
}
}
if (downHostList.size() > 0)
queueNotification(new ClusterManagerMessage(ClusterManagerMessage.MessageType.nodeRemoved, downHostList));
} else {
s_logger.info("No inactive management server node found");
}
}
private void peerScan() throws ActiveFencingException {
Date cutTime = DateUtil.currentGMTTime();
Profiler profiler = new Profiler();
profiler.start();
Profiler profilerQueryActiveList = new Profiler();
profilerQueryActiveList.start();
List<ManagementServerHostVO> currentList = _mshostDao.getActiveList(new Date(cutTime.getTime() - HeartbeatThreshold.value()));
profilerQueryActiveList.stop();
Profiler profilerSyncClusterInfo = new Profiler();
profilerSyncClusterInfo.start();
List<ManagementServerHostVO> removedNodeList = new ArrayList<ManagementServerHostVO>();
List<ManagementServerHostVO> invalidatedNodeList = new ArrayList<ManagementServerHostVO>();
if (_mshostId != null) {
if (_mshostPeerDao.countStateSeenInPeers(_mshostId, _runId, ManagementServerHost.State.Down) > 0) {
String msg =
"We have detected that at least one management server peer reports that this management server is down, perform active fencing to avoid split-brain situation";
s_logger.error(msg);
throw new ActiveFencingException(msg);
}
// only if we have already attached to cluster, will we start to check leaving nodes
for (Map.Entry<Long, ManagementServerHostVO> entry : _activePeers.entrySet()) {
ManagementServerHostVO current = getInListById(entry.getKey(), currentList);
if (current == null) {
if (entry.getKey().longValue() != _mshostId.longValue()) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Detected management node left, id:" + entry.getKey() + ", nodeIP:" + entry.getValue().getServiceIP());
}
removedNodeList.add(entry.getValue());
}
} else {
if (current.getRunid() == 0) {
if (entry.getKey().longValue() != _mshostId.longValue()) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Detected management node left because of invalidated session, id:" + entry.getKey() + ", nodeIP:" +
entry.getValue().getServiceIP());
}
invalidatedNodeList.add(entry.getValue());
}
} else {
if (entry.getValue().getRunid() != current.getRunid()) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Detected management node left and rejoined quickly, id:" + entry.getKey() + ", nodeIP:" + entry.getValue().getServiceIP());
}
entry.getValue().setRunid(current.getRunid());
}
}
}
}
}
profilerSyncClusterInfo.stop();
Profiler profilerInvalidatedNodeList = new Profiler();
profilerInvalidatedNodeList.start();
// process invalidated node list
if (invalidatedNodeList.size() > 0) {
for (ManagementServerHostVO mshost : invalidatedNodeList) {
_activePeers.remove(mshost.getId());
try {
JmxUtil.unregisterMBean("ClusterManager", "Node " + mshost.getId());
} catch (Exception e) {
s_logger.warn("Unable to deregiester cluster node from JMX monitoring due to exception " + e.toString());
}
}
queueNotification(new ClusterManagerMessage(ClusterManagerMessage.MessageType.nodeRemoved, invalidatedNodeList));
}
profilerInvalidatedNodeList.stop();
Profiler profilerRemovedList = new Profiler();
profilerRemovedList.start();
// process removed node list
Iterator<ManagementServerHostVO> it = removedNodeList.iterator();
while (it.hasNext()) {
ManagementServerHostVO mshost = it.next();
if (!pingManagementNode(mshost)) {
s_logger.warn("Management node " + mshost.getId() + " is detected inactive by timestamp and also not pingable");
_activePeers.remove(mshost.getId());
try {
JmxUtil.unregisterMBean("ClusterManager", "Node " + mshost.getId());
} catch (Exception e) {
s_logger.warn("Unable to deregiester cluster node from JMX monitoring due to exception " + e.toString());
}
} else {
s_logger.info("Management node " + mshost.getId() + " is detected inactive by timestamp but is pingable");
it.remove();
}
}
if (removedNodeList.size() > 0) {
queueNotification(new ClusterManagerMessage(ClusterManagerMessage.MessageType.nodeRemoved, removedNodeList));
}
profilerRemovedList.stop();
List<ManagementServerHostVO> newNodeList = new ArrayList<ManagementServerHostVO>();
for (ManagementServerHostVO mshost : currentList) {
if (!_activePeers.containsKey(mshost.getId())) {
_activePeers.put(mshost.getId(), mshost);
if (s_logger.isDebugEnabled()) {
s_logger.debug("Detected management node joined, id:" + mshost.getId() + ", nodeIP:" + mshost.getServiceIP());
}
newNodeList.add(mshost);
try {
JmxUtil.registerMBean("ClusterManager", "Node " + mshost.getId(), new ClusterManagerMBeanImpl(this, mshost));
} catch (Exception e) {
s_logger.warn("Unable to regiester cluster node into JMX monitoring due to exception " + ExceptionUtil.toString(e));
}
}
}
if (newNodeList.size() > 0) {
queueNotification(new ClusterManagerMessage(ClusterManagerMessage.MessageType.nodeAdded, newNodeList));
}
profiler.stop();
if (profiler.getDuration() >= HeartbeatInterval.value()) {
if (s_logger.isDebugEnabled())
s_logger.debug("Peer scan takes too long to finish. profiler: " + profiler.toString() + ", profilerQueryActiveList: " +
profilerQueryActiveList.toString() + ", profilerSyncClusterInfo: " + profilerSyncClusterInfo.toString() + ", profilerInvalidatedNodeList: " +
profilerInvalidatedNodeList.toString() + ", profilerRemovedList: " + profilerRemovedList.toString());
}
}
private static ManagementServerHostVO getInListById(Long id, List<ManagementServerHostVO> l) {
for (ManagementServerHostVO mshost : l) {
if (mshost.getId() == id) {
return mshost;
}
}
return null;
}
@Override
@DB
public boolean start() {
if (s_logger.isInfoEnabled()) {
s_logger.info("Starting cluster manager, msid : " + _msId);
}
ManagementServerHostVO mshost = Transaction.execute(new TransactionCallback<ManagementServerHostVO>() {
@Override
public ManagementServerHostVO doInTransaction(TransactionStatus status) {
final Class<?> c = this.getClass();
String version = c.getPackage().getImplementationVersion();
ManagementServerHostVO mshost = _mshostDao.findByMsid(_msId);
if (mshost == null) {
mshost = new ManagementServerHostVO();
mshost.setMsid(_msId);
mshost.setRunid(getCurrentRunId());
mshost.setName(NetUtils.getHostName());
mshost.setVersion(version);
mshost.setServiceIP(_clusterNodeIP);
mshost.setServicePort(_currentServiceAdapter.getServicePort());
mshost.setLastUpdateTime(DateUtil.currentGMTTime());
mshost.setRemoved(null);
mshost.setAlertCount(0);
mshost.setState(ManagementServerHost.State.Up);
_mshostDao.persist(mshost);
if (s_logger.isInfoEnabled()) {
s_logger.info("New instance of management server msid " + _msId + " is being started");
}
} else {
if (s_logger.isInfoEnabled()) {
s_logger.info("Management server " + _msId + " is being started");
}
_mshostDao.update(mshost.getId(), getCurrentRunId(), NetUtils.getHostName(), version, _clusterNodeIP, _currentServiceAdapter.getServicePort(),
DateUtil.currentGMTTime());
}
return mshost;
}
});
_mshostId = mshost.getId();
if (s_logger.isInfoEnabled()) {
s_logger.info("Management server (host id : " + _mshostId + ") is being started at " + _clusterNodeIP + ":" + _currentServiceAdapter.getServicePort());
}
_mshostPeerDao.clearPeerInfo(_mshostId);
// use seperate thread for heartbeat updates
_heartbeatScheduler.scheduleAtFixedRate(getHeartbeatTask(), HeartbeatInterval.value(), HeartbeatInterval.value(), TimeUnit.MILLISECONDS);
_notificationExecutor.submit(getNotificationTask());
if (s_logger.isInfoEnabled()) {
s_logger.info("Cluster manager was started successfully");
}
return true;
}
@Override
@DB
public boolean stop() {
if (_mshostId != null) {
ManagementServerHostVO mshost = _mshostDao.findByMsid(_msId);
mshost.setState(ManagementServerHost.State.Down);
_mshostDao.update(_mshostId, mshost);
}
_heartbeatScheduler.shutdownNow();
_executor.shutdownNow();
try {
_heartbeatScheduler.awaitTermination(EXECUTOR_SHUTDOWN_TIMEOUT, TimeUnit.MILLISECONDS);
_executor.awaitTermination(EXECUTOR_SHUTDOWN_TIMEOUT, TimeUnit.MILLISECONDS);
} catch (InterruptedException e) {
}
if (s_logger.isInfoEnabled()) {
s_logger.info("Cluster manager is stopped");
}
return true;
}
@Override
public boolean configure(String name, Map<String, Object> params) throws ConfigurationException {
if (s_logger.isInfoEnabled()) {
s_logger.info("Start configuring cluster manager : " + name);
}
Properties dbProps = DbProperties.getDbProperties();
_clusterNodeIP = dbProps.getProperty("cluster.node.IP");
if (_clusterNodeIP == null) {
_clusterNodeIP = "127.0.0.1";
}
_clusterNodeIP = _clusterNodeIP.trim();
if (s_logger.isInfoEnabled()) {
s_logger.info("Cluster node IP : " + _clusterNodeIP);
}
if (!NetUtils.isLocalAddress(_clusterNodeIP)) {
throw new ConfigurationException("cluster node IP should be valid local address where the server is running, please check your configuration");
}
for (int i = 0; i < DEFAULT_OUTGOING_WORKERS; i++)
_executor.execute(getClusterPduSendingTask());
// notification task itself in turn works as a task dispatcher
_executor.execute(getClusterPduNotificationTask());
if (_serviceAdapters == null) {
throw new ConfigurationException("Unable to get cluster service adapters");
}
_currentServiceAdapter = _serviceAdapters.get(0);
if (_currentServiceAdapter == null) {
throw new ConfigurationException("Unable to set current cluster service adapter");
}
checkConflicts();
if (s_logger.isInfoEnabled()) {
s_logger.info("Cluster manager is configured.");
}
return true;
}
@Override
public long getManagementNodeId() {
return _msId;
}
@Override
public long getCurrentRunId() {
return _runId;
}
public boolean isManagementNodeAlive(long msid) {
ManagementServerHostVO mshost = _mshostDao.findByMsid(msid);
if (mshost != null) {
if (mshost.getLastUpdateTime().getTime() >= DateUtil.currentGMTTime().getTime() - HeartbeatThreshold.value()) {
return true;
}
}
return false;
}
public boolean pingManagementNode(long msid) {
ManagementServerHostVO mshost = _mshostDao.findByMsid(msid);
if (mshost == null) {
return false;
}
return pingManagementNode(mshost);
}
@Override
public String getConfigComponentName() {
return ClusterManager.class.getSimpleName();
}
@Override
public ConfigKey<?>[] getConfigKeys() {
return new ConfigKey<?>[] {HeartbeatInterval, HeartbeatThreshold};
}
private boolean pingManagementNode(ManagementServerHostVO mshost) {
String targetIp = mshost.getServiceIP();
if ("127.0.0.1".equals(targetIp) || "0.0.0.0".equals(targetIp)) {
s_logger.info("ping management node cluster service can not be performed on self");
return false;
}
int retry = 10;
while (--retry > 0) {
SocketChannel sch = null;
try {
s_logger.info("Trying to connect to " + targetIp);
sch = SocketChannel.open();
sch.configureBlocking(true);
sch.socket().setSoTimeout(5000);
InetSocketAddress addr = new InetSocketAddress(targetIp, mshost.getServicePort());
sch.connect(addr);
return true;
} catch (IOException e) {
if (e instanceof ConnectException) {
s_logger.error("Unable to ping management server at " + targetIp + ":" + mshost.getServicePort() + " due to ConnectException", e);
return false;
}
} finally {
if (sch != null) {
try {
sch.close();
} catch (IOException e) {
}
}
}
try {
Thread.sleep(1000);
} catch (InterruptedException ex) {
}
}
s_logger.error("Unable to ping management server at " + targetIp + ":" + mshost.getServicePort() + " after retries");
return false;
}
public int getHeartbeatInterval() {
return HeartbeatInterval.value();
}
private void checkConflicts() throws ConfigurationException {
Date cutTime = DateUtil.currentGMTTime();
List<ManagementServerHostVO> peers = _mshostDao.getActiveList(new Date(cutTime.getTime() - HeartbeatThreshold.value()));
for (ManagementServerHostVO peer : peers) {
String peerIP = peer.getServiceIP().trim();
if (_clusterNodeIP.equals(peerIP)) {
if ("127.0.0.1".equals(_clusterNodeIP)) {
if (pingManagementNode(peer.getMsid())) {
String msg = "Detected another management node with localhost IP is already running, please check your cluster configuration";
s_logger.error(msg);
throw new ConfigurationException(msg);
} else {
String msg =
"Detected another management node with localhost IP is considered as running in DB, however it is not pingable, we will continue cluster initialization with this management server node";
s_logger.info(msg);
}
} else {
if (pingManagementNode(peer.getMsid())) {
String msg =
"Detected that another management node with the same IP " + peer.getServiceIP() +
" is already running, please check your cluster configuration";
s_logger.error(msg);
throw new ConfigurationException(msg);
} else {
String msg =
"Detected that another management node with the same IP " + peer.getServiceIP() +
" is considered as running in DB, however it is not pingable, we will continue cluster initialization with this management server node";
s_logger.info(msg);
}
}
}
}
}
}
Reference in New Issue View Git Blame Copy Permalink