Examples with ClusterTopologyCheckedException org.apache.ignite.internal.cluster.ClusterTopologyCheckedException used on opensource projects

Example 36 with ClusterTopologyCheckedException

use of org.apache.ignite.internal.cluster.ClusterTopologyCheckedException in project ignite by apache.

the class CacheContinuousQueryHandler method register.

/**
 * {@inheritDoc}
 */
@Override
public RegisterStatus register(final UUID nodeId, final UUID routineId, final GridKernalContext ctx) throws IgniteCheckedException {
    assert nodeId != null;
    assert routineId != null;
    assert ctx != null;
    initLocalListener(locLsnr, ctx);
    if (initFut == null) {
        initFut = p2pUnmarshalFut.chain((fut) -> {
            try {
                fut.get();
                initRemoteFilter(getEventFilter0(), ctx);
                IgniteClosure trans = getTransformer0();
                if (trans != null)
                    ctx.resource().injectGeneric(trans);
            } catch (IgniteCheckedException | ExceptionInInitializerError e) {
                throw new IgniteException("Failed to initialize a continuous query.", e);
            }
            return null;
        });
    }
    if (initFut.error() != null)
        throw new IgniteCheckedException("Failed to initialize a continuous query.", initFut.error());
    entryBufs = new ConcurrentHashMap<>();
    ackBuf = new CacheContinuousQueryAcknowledgeBuffer();
    rcvs = new ConcurrentHashMap<>();
    this.nodeId = nodeId;
    this.routineId = routineId;
    this.ctx = ctx;
    final boolean loc = nodeId.equals(ctx.localNodeId());
    assert !skipPrimaryCheck || loc;
    log = ctx.log(CU.CONTINUOUS_QRY_LOG_CATEGORY);
    CacheContinuousQueryListener<K, V> lsnr = new CacheContinuousQueryListener<K, V>() {

        @Override
        public void onBeforeRegister() {
            GridCacheContext<K, V> cctx = cacheContext(ctx);
            if (cctx != null && !cctx.isLocal())
                cctx.topology().readLock();
        }

        @Override
        public void onAfterRegister() {
            GridCacheContext<K, V> cctx = cacheContext(ctx);
            if (cctx != null && !cctx.isLocal())
                cctx.topology().readUnlock();
        }

        @Override
        public void onRegister() {
            GridCacheContext<K, V> cctx = cacheContext(ctx);
            if (cctx != null && !cctx.isLocal())
                locInitUpdCntrs = toCountersMap(cctx.topology().localUpdateCounters(false));
        }

        @Override
        public boolean keepBinary() {
            return keepBinary;
        }

        @Override
        public void onEntryUpdated(final CacheContinuousQueryEvent<K, V> evt, boolean primary, final boolean recordIgniteEvt, GridDhtAtomicAbstractUpdateFuture fut) {
            if (ignoreExpired && evt.getEventType() == EventType.EXPIRED)
                return;
            if (log.isDebugEnabled())
                log.debug("Entry updated on affinity node [evt=" + evt + ", primary=" + primary + ']');
            final GridCacheContext<K, V> cctx = cacheContext(ctx);
            // Check that cache stopped.
            if (cctx == null)
                return;
            if (!needNotify(false, cctx, -1, -1, evt))
                return;
            // skipPrimaryCheck is set only when listen locally for replicated cache events.
            assert !skipPrimaryCheck || (cctx.isReplicated() && ctx.localNodeId().equals(nodeId));
            if (asyncCb) {
                ContinuousQueryAsyncClosure clsr = new ContinuousQueryAsyncClosure(primary, evt, recordIgniteEvt, fut);
                ctx.pools().asyncCallbackPool().execute(clsr, evt.partitionId());
            } else {
                final boolean notify = filter(evt);
                if (log.isDebugEnabled())
                    log.debug("Filter invoked for event [evt=" + evt + ", primary=" + primary + ", notify=" + notify + ']');
                if (primary || skipPrimaryCheck)
                    onEntryUpdate(evt, notify, loc, recordIgniteEvt);
                else
                    handleBackupEntry(cctx, evt.entry());
            }
        }

        @Override
        public void onUnregister() {
            try {
                CacheEntryEventFilter filter = getEventFilter();
                if (filter instanceof PlatformContinuousQueryFilter)
                    ((PlatformContinuousQueryFilter) filter).onQueryUnregister();
            } catch (IgniteCheckedException e) {
                if (log.isDebugEnabled()) {
                    log.debug("Failed to execute the onUnregister callback " + "on the continuoue query listener. " + "[nodeId=" + nodeId + ", routineId=" + routineId + ", cacheName=" + cacheName + ", err=" + e + "]");
                }
            }
        }

        @Override
        public void cleanupOnAck(Map<Integer, Long> updateCntrs) {
            for (Map.Entry<Integer, Long> e : updateCntrs.entrySet()) {
                CacheContinuousQueryEventBuffer buf = entryBufs.get(e.getKey());
                if (buf != null)
                    buf.cleanupOnAck(e.getValue());
            }
        }

        @Override
        public void flushOnExchangeDone(GridKernalContext ctx, AffinityTopologyVersion topVer) {
            assert topVer != null;
            try {
                GridCacheContext<K, V> cctx = cacheContext(ctx);
                ClusterNode node = ctx.discovery().node(nodeId);
                for (Map.Entry<Integer, CacheContinuousQueryEventBuffer> bufE : entryBufs.entrySet()) {
                    CacheContinuousQueryEventBuffer buf = bufE.getValue();
                    Collection<CacheContinuousQueryEntry> entries = buf.flushOnExchange((cntr, filtered) -> createFilteredEntry(cctx.cacheId(), bufE.getKey(), topVer, cntr, filtered));
                    if (entries == null || node == null)
                        continue;
                    for (CacheContinuousQueryEntry e : entries) {
                        e.markBackup();
                        if (!e.isFiltered())
                            prepareEntry(cctx, nodeId, e);
                    }
                    ctx.continuous().addBackupNotification(nodeId, routineId, entries, topic);
                }
            } catch (IgniteCheckedException e) {
                U.error(ctx.log(CU.CONTINUOUS_QRY_LOG_CATEGORY), "Failed to send backup event notification to node: " + nodeId, e);
            }
        }

        @Override
        public void acknowledgeBackupOnTimeout(GridKernalContext ctx) {
            sendBackupAcknowledge(ackBuf.acknowledgeOnTimeout(), routineId, ctx);
        }

        @Override
        public void skipUpdateEvent(CacheContinuousQueryEvent<K, V> evt, AffinityTopologyVersion topVer, boolean primary) {
            assert evt != null;
            CacheContinuousQueryEntry e = evt.entry();
            e.markFiltered();
            onEntryUpdated(evt, primary, false, null);
        }

        @Override
        public CounterSkipContext skipUpdateCounter(final GridCacheContext cctx, @Nullable CounterSkipContext skipCtx, int part, long cntr, AffinityTopologyVersion topVer, boolean primary) {
            if (skipCtx == null)
                skipCtx = new CounterSkipContext(part, cntr, topVer);
            if (!needNotify(true, cctx, part, cntr, null))
                return skipCtx;
            if (loc) {
                assert !locOnly;
                final Collection<CacheEntryEvent<? extends K, ? extends V>> evts = handleEvent(ctx, skipCtx.entry());
                if (!evts.isEmpty()) {
                    if (asyncCb) {
                        ctx.pools().asyncCallbackPool().execute(new Runnable() {

                            @Override
                            public void run() {
                                try {
                                    notifyLocalListener(evts, getTransformer());
                                } catch (IgniteCheckedException ex) {
                                    U.error(ctx.log(CU.CONTINUOUS_QRY_LOG_CATEGORY), "Failed to notify local listener.", ex);
                                }
                            }
                        }, part);
                    } else
                        skipCtx.addProcessClosure(new Runnable() {

                            @Override
                            public void run() {
                                try {
                                    notifyLocalListener(evts, getTransformer());
                                } catch (IgniteCheckedException ex) {
                                    U.error(ctx.log(CU.CONTINUOUS_QRY_LOG_CATEGORY), "Failed to notify local listener.", ex);
                                }
                            }
                        });
                }
                return skipCtx;
            }
            CacheContinuousQueryEventBuffer buf = partitionBuffer(cctx, part);
            final Object entryOrList = buf.processEntry(skipCtx.entry(), !primary);
            if (entryOrList != null) {
                skipCtx.addProcessClosure(new Runnable() {

                    @Override
                    public void run() {
                        try {
                            ctx.continuous().addNotification(nodeId, routineId, entryOrList, topic, false, true);
                        } catch (ClusterTopologyCheckedException ex) {
                            if (log.isDebugEnabled())
                                log.debug("Failed to send event notification to node, node left cluster " + "[node=" + nodeId + ", err=" + ex + ']');
                        } catch (IgniteCheckedException ex) {
                            U.error(ctx.log(CU.CONTINUOUS_QRY_LOG_CATEGORY), "Failed to send event notification to node: " + nodeId, ex);
                        }
                    }
                });
            }
            return skipCtx;
        }

        @Override
        public void onPartitionEvicted(int part) {
            entryBufs.remove(part);
        }

        @Override
        public boolean oldValueRequired() {
            return oldValRequired;
        }

        @Override
        public boolean notifyExisting() {
            return notifyExisting;
        }

        private String taskName() {
            return ctx.security().enabled() ? ctx.task().resolveTaskName(taskHash) : null;
        }

        @Override
        public boolean isPrimaryOnly() {
            return locOnly && !skipPrimaryCheck;
        }

        /**
         * Checks whether it is need to notify listeners.
         *
         * @param skipEvt {@code True} if this is a skip counter event.
         * @param cctx Cache context.
         * @param part Partition id.
         * @param cntr Update counter.
         * @param evt CQ event.
         * @return {@code True} if notification should happen immediately, or {@code false} if it should be delayed.
         */
        private boolean needNotify(boolean skipEvt, GridCacheContext cctx, int part, long cntr, CacheContinuousQueryEvent evt) {
            assert !skipEvt || evt == null;
            // part == -1 && cntr == -1 means skip counter.
            assert skipEvt || part == -1 && cntr == -1;
            if (!cctx.mvccEnabled() || cctx.isLocal())
                return true;
            assert locInitUpdCntrs != null;
            cntr = skipEvt ? cntr : evt.getPartitionUpdateCounter();
            part = skipEvt ? part : evt.partitionId();
            T2<Long, Long> initCntr = locInitUpdCntrs.get(part);
            // Do not notify listener if entry was updated before the query is started.
            return initCntr == null || cntr >= initCntr.get2();
        }
    };
    CacheContinuousQueryManager mgr = manager(ctx);
    if (mgr == null)
        return RegisterStatus.DELAYED;
    RegisterStatus regStatus = mgr.registerListener(routineId, lsnr, internal);
    if (regStatus == RegisterStatus.REGISTERED)
        initFut.listen(res -> sendQueryExecutedEvent());
    return regStatus;
}

Example 37 with ClusterTopologyCheckedException

use of org.apache.ignite.internal.cluster.ClusterTopologyCheckedException in project ignite by apache.

the class GridCacheDistributedQueryManager method sendQueryResponse.

/**
 * Sends cache query response.
 *
 * @param nodeId Node to send response.
 * @param res Cache query response.
 * @param timeout Message timeout.
 * @return {@code true} if response was sent, {@code false} otherwise.
 */
private boolean sendQueryResponse(UUID nodeId, GridCacheQueryResponse res, long timeout) {
    ClusterNode node = cctx.node(nodeId);
    if (node == null)
        return false;
    int attempt = 1;
    IgniteCheckedException err = null;
    while (!Thread.currentThread().isInterrupted()) {
        try {
            if (log.isDebugEnabled())
                log.debug("Send query response: " + res);
            Object topic = topic(nodeId, res.requestId());
            cctx.io().sendOrderedMessage(node, topic, res, GridIoPolicy.QUERY_POOL, timeout > 0 ? timeout : Long.MAX_VALUE);
            return true;
        } catch (ClusterTopologyCheckedException ignored) {
            if (log.isDebugEnabled())
                log.debug("Failed to send query response since node left grid [nodeId=" + nodeId + ", res=" + res + "]");
            return false;
        } catch (IgniteCheckedException e) {
            if (err == null)
                err = e;
            if (Thread.currentThread().isInterrupted())
                break;
            if (attempt < RESEND_ATTEMPTS) {
                if (log.isDebugEnabled())
                    log.debug("Failed to send queries response (will try again) [nodeId=" + nodeId + ", res=" + res + ", attempt=" + attempt + ", err=" + e + "]");
                if (!Thread.currentThread().isInterrupted())
                    try {
                        U.sleep(RESEND_FREQ);
                    } catch (IgniteInterruptedCheckedException e1) {
                        U.error(log, "Waiting for queries response resending was interrupted (response will not be sent) " + "[nodeId=" + nodeId + ", response=" + res + "]", e1);
                        return false;
                    }
            } else {
                U.error(log, "Failed to sender cache response [nodeId=" + nodeId + ", response=" + res + "]", err);
                return false;
            }
        }
        attempt++;
    }
    return false;
}

Example 38 with ClusterTopologyCheckedException

use of org.apache.ignite.internal.cluster.ClusterTopologyCheckedException in project ignite by apache.

the class ClusterProcessor method initDiagnosticListeners.

/**
 * @throws IgniteCheckedException If failed.
 */
public void initDiagnosticListeners() throws IgniteCheckedException {
    ctx.event().addLocalEventListener(new GridLocalEventListener() {

        @Override
        public void onEvent(Event evt) {
            assert evt instanceof DiscoveryEvent;
            assert evt.type() == EVT_NODE_FAILED || evt.type() == EVT_NODE_LEFT;
            DiscoveryEvent discoEvt = (DiscoveryEvent) evt;
            UUID nodeId = discoEvt.eventNode().id();
            ConcurrentHashMap<Long, InternalDiagnosticFuture> futs = diagnosticFutMap.get();
            if (futs != null) {
                for (InternalDiagnosticFuture fut : futs.values()) {
                    if (fut.nodeId.equals(nodeId))
                        fut.onDone(new IgniteDiagnosticInfo("Target node failed: " + nodeId));
                }
            }
            allNodesMetrics.remove(nodeId);
        }
    }, EVT_NODE_FAILED, EVT_NODE_LEFT);
    ctx.io().addMessageListener(TOPIC_INTERNAL_DIAGNOSTIC, new GridMessageListener() {

        @Override
        public void onMessage(UUID nodeId, Object msg, byte plc) {
            if (msg instanceof IgniteDiagnosticMessage) {
                IgniteDiagnosticMessage msg0 = (IgniteDiagnosticMessage) msg;
                if (msg0.request()) {
                    ClusterNode node = ctx.discovery().node(nodeId);
                    if (node == null) {
                        if (diagnosticLog.isDebugEnabled()) {
                            diagnosticLog.debug("Skip diagnostic request, sender node left " + "[node=" + nodeId + ", msg=" + msg + ']');
                        }
                        return;
                    }
                    byte[] diagRes;
                    IgniteClosure<GridKernalContext, IgniteDiagnosticInfo> c;
                    try {
                        c = msg0.unmarshal(marsh);
                        diagRes = marsh.marshal(c.apply(ctx));
                    } catch (Exception e) {
                        U.error(diagnosticLog, "Failed to run diagnostic closure: " + e, e);
                        try {
                            IgniteDiagnosticInfo errInfo = new IgniteDiagnosticInfo("Failed to run diagnostic closure: " + e);
                            diagRes = marsh.marshal(errInfo);
                        } catch (Exception e0) {
                            U.error(diagnosticLog, "Failed to marshal diagnostic closure result: " + e, e);
                            diagRes = null;
                        }
                    }
                    IgniteDiagnosticMessage res = IgniteDiagnosticMessage.createResponse(diagRes, msg0.futureId());
                    try {
                        ctx.io().sendToGridTopic(node, TOPIC_INTERNAL_DIAGNOSTIC, res, GridIoPolicy.SYSTEM_POOL);
                    } catch (ClusterTopologyCheckedException e) {
                        if (diagnosticLog.isDebugEnabled()) {
                            diagnosticLog.debug("Failed to send diagnostic response, node left " + "[node=" + nodeId + ", msg=" + msg + ']');
                        }
                    } catch (IgniteCheckedException e) {
                        U.error(diagnosticLog, "Failed to send diagnostic response [msg=" + msg0 + "]", e);
                    }
                } else {
                    InternalDiagnosticFuture fut = diagnosticFuturesMap().get(msg0.futureId());
                    if (fut != null) {
                        IgniteDiagnosticInfo res;
                        try {
                            res = msg0.unmarshal(marsh);
                            if (res == null)
                                res = new IgniteDiagnosticInfo("Remote node failed to marshal response.");
                        } catch (Exception e) {
                            U.error(diagnosticLog, "Failed to unmarshal diagnostic response: " + e, e);
                            res = new IgniteDiagnosticInfo("Failed to unmarshal diagnostic response: " + e);
                        }
                        fut.onResponse(res);
                    } else
                        U.warn(diagnosticLog, "Failed to find diagnostic message future [msg=" + msg0 + ']');
                }
            } else
                U.warn(diagnosticLog, "Received unexpected message: " + msg);
        }
    });
    if (sndMetrics) {
        ctx.io().addMessageListener(TOPIC_METRICS, new GridMessageListener() {

            @Override
            public void onMessage(UUID nodeId, Object msg, byte plc) {
                if (msg instanceof ClusterMetricsUpdateMessage)
                    processMetricsUpdateMessage(nodeId, (ClusterMetricsUpdateMessage) msg);
                else
                    U.warn(log, "Received unexpected message for TOPIC_METRICS: " + msg);
            }
        });
    }
}

Example 39 with ClusterTopologyCheckedException

use of org.apache.ignite.internal.cluster.ClusterTopologyCheckedException in project ignite by apache.

the class ClusterProcessor method updateMetrics.

/**
 */
private void updateMetrics() {
    if (ctx.isStopping() || ctx.clientDisconnected())
        return;
    ClusterNode oldest = ctx.discovery().oldestAliveServerNode(AffinityTopologyVersion.NONE);
    if (oldest == null)
        return;
    if (ctx.localNodeId().equals(oldest.id())) {
        IgniteClusterNode locNode = (IgniteClusterNode) ctx.discovery().localNode();
        locNode.setMetrics(metricsProvider.metrics());
        locNode.setCacheMetrics(metricsProvider.cacheMetrics());
        ClusterNodeMetrics metrics = new ClusterNodeMetrics(locNode.metrics(), locNode.cacheMetrics());
        try {
            byte[] metricsBytes = U.zip(U.marshal(ctx.config().getMarshaller(), metrics));
            allNodesMetrics.put(ctx.localNodeId(), metricsBytes);
        } catch (IgniteCheckedException e) {
            U.warn(log, "Failed to marshal local node metrics: " + e, e);
        }
        ctx.discovery().metricsUpdateEvent(ctx.discovery().discoCache(), locNode);
        Collection<ClusterNode> allNodes = ctx.discovery().allNodes();
        ClusterMetricsUpdateMessage msg = new ClusterMetricsUpdateMessage(new HashMap<>(allNodesMetrics));
        for (ClusterNode node : allNodes) {
            if (ctx.localNodeId().equals(node.id()) || !ctx.discovery().alive(node.id()))
                continue;
            try {
                ctx.io().sendToGridTopic(node, TOPIC_METRICS, msg, GridIoPolicy.SYSTEM_POOL);
            } catch (ClusterTopologyCheckedException e) {
                if (log.isDebugEnabled())
                    log.debug("Failed to send metrics update, node failed: " + e);
            } catch (IgniteCheckedException e) {
                U.warn(log, "Failed to send metrics update: " + e, e);
            }
        }
    } else {
        ClusterNodeMetrics metrics = new ClusterNodeMetrics(metricsProvider.metrics(), metricsProvider.cacheMetrics());
        try {
            byte[] metricsBytes = U.zip(U.marshal(ctx.config().getMarshaller(), metrics));
            ClusterMetricsUpdateMessage msg = new ClusterMetricsUpdateMessage(metricsBytes);
            ctx.io().sendToGridTopic(oldest, TOPIC_METRICS, msg, GridIoPolicy.SYSTEM_POOL);
        } catch (ClusterTopologyCheckedException e) {
            if (log.isDebugEnabled())
                log.debug("Failed to send metrics update to oldest, node failed: " + e);
        } catch (IgniteCheckedException e) {
            LT.warn(log, e, "Failed to send metrics update to oldest: " + e, false, false);
        }
    }
}

Example 40 with ClusterTopologyCheckedException

use of org.apache.ignite.internal.cluster.ClusterTopologyCheckedException in project ignite by apache.

the class GridJobProcessor method handleException.

/**
 * Handles errors that happened prior to job creation.
 *
 * @param node Sender node.
 * @param req Job execution request.
 * @param ex Exception that happened.
 * @param endTime Job end time.
 */
private void handleException(ClusterNode node, GridJobExecuteRequest req, IgniteException ex, long endTime) {
    UUID locNodeId = ctx.localNodeId();
    ClusterNode sndNode = ctx.discovery().node(node.id());
    if (sndNode == null) {
        U.warn(log, "Failed to reply to sender node because it left grid [nodeId=" + node.id() + ", jobId=" + req.getJobId() + ']');
        if (ctx.event().isRecordable(EVT_JOB_FAILED)) {
            JobEvent evt = new JobEvent();
            evt.jobId(req.getJobId());
            evt.message("Job reply failed (original task node left grid): " + req.getJobId());
            evt.node(ctx.discovery().localNode());
            evt.taskName(req.getTaskName());
            evt.taskClassName(req.getTaskClassName());
            evt.taskSessionId(req.getSessionId());
            evt.type(EVT_JOB_FAILED);
            evt.taskNode(node);
            evt.taskSubjectId(securitySubjectId(ctx));
            // Record job reply failure.
            ctx.event().record(evt);
        }
        return;
    }
    try {
        boolean loc = ctx.localNodeId().equals(sndNode.id()) && !ctx.config().isMarshalLocalJobs();
        GridJobExecuteResponse jobRes = new GridJobExecuteResponse(locNodeId, req.getSessionId(), req.getJobId(), loc ? null : U.marshal(marsh, ex), ex, loc ? null : U.marshal(marsh, null), null, loc ? null : U.marshal(marsh, null), null, false, null);
        if (req.isSessionFullSupport()) {
            // Send response to designated job topic.
            // Always go through communication to preserve order,
            // if attributes are enabled.
            // Job response topic.
            Object topic = TOPIC_TASK.topic(req.getJobId(), locNodeId);
            long timeout = endTime - U.currentTimeMillis();
            if (timeout <= 0)
                // Ignore the actual timeout and send response anyway.
                timeout = 1;
            // Send response to designated job topic.
            // Always go through communication to preserve order.
            ctx.io().sendOrderedMessage(sndNode, topic, jobRes, req.isInternal() ? MANAGEMENT_POOL : SYSTEM_POOL, timeout, false);
        } else if (ctx.localNodeId().equals(sndNode.id()))
            ctx.task().processJobExecuteResponse(ctx.localNodeId(), jobRes);
        else
            // Send response to common topic as unordered message.
            ctx.io().sendToGridTopic(sndNode, TOPIC_TASK, jobRes, req.isInternal() ? MANAGEMENT_POOL : SYSTEM_POOL);
    } catch (IgniteCheckedException e) {
        // The only option here is to log, as we must assume that resending will fail too.
        if ((e instanceof ClusterTopologyCheckedException) || isDeadNode(node.id()))
            // Avoid stack trace for left nodes.
            U.error(log, "Failed to reply to sender node because it left grid [nodeId=" + node.id() + ", jobId=" + req.getJobId() + ']');
        else {
            assert sndNode != null;
            U.error(log, "Error sending reply for job [nodeId=" + sndNode.id() + ", jobId=" + req.getJobId() + ']', e);
        }
        if (ctx.event().isRecordable(EVT_JOB_FAILED)) {
            JobEvent evt = new JobEvent();
            evt.jobId(req.getJobId());
            evt.message("Failed to send reply for job: " + req.getJobId());
            evt.node(ctx.discovery().localNode());
            evt.taskName(req.getTaskName());
            evt.taskClassName(req.getTaskClassName());
            evt.taskSessionId(req.getSessionId());
            evt.type(EVT_JOB_FAILED);
            evt.taskNode(node);
            evt.taskSubjectId(securitySubjectId(ctx));
            // Record job reply failure.
            ctx.event().record(evt);
        }
    }
}