Examples with ClusterTopologyException org.apache.ignite.cluster.ClusterTopologyException used on opensource projects

Example 1 with ClusterTopologyException

use of org.apache.ignite.cluster.ClusterTopologyException in project ignite by apache.

the class DataStreamerImpl method load0.

/**
     * @param entries Entries.
     * @param resFut Result future.
     * @param activeKeys Active keys.
     * @param remaps Remaps count.
     */
private void load0(Collection<? extends DataStreamerEntry> entries, final GridFutureAdapter<Object> resFut, @Nullable final Collection<KeyCacheObjectWrapper> activeKeys, final int remaps) {
    try {
        assert entries != null;
        final boolean remap = remaps > 0;
        if (!remap) {
            // Failed data should be processed prior to new data.
            acquireRemapSemaphore();
        }
        if (!isWarningPrinted) {
            synchronized (this) {
                if (!allowOverwrite() && !isWarningPrinted) {
                    U.warn(log, "Data streamer will not overwrite existing cache entries for better performance " + "(to change, set allowOverwrite to true)");
                }
                isWarningPrinted = true;
            }
        }
        Map<ClusterNode, Collection<DataStreamerEntry>> mappings = new HashMap<>();
        boolean initPda = ctx.deploy().enabled() && jobPda == null;
        GridCacheAdapter cache = ctx.cache().internalCache(cacheName);
        if (cache == null)
            throw new IgniteCheckedException("Cache not created or already destroyed.");
        GridCacheContext cctx = cache.context();
        GridCacheGateway gate = null;
        if (!allowOverwrite() && !cctx.isLocal()) {
            // Cases where cctx required.
            gate = cctx.gate();
            gate.enter();
        }
        try {
            AffinityTopologyVersion topVer = allowOverwrite() || cctx.isLocal() ? ctx.cache().context().exchange().readyAffinityVersion() : cctx.topology().topologyVersion();
            for (DataStreamerEntry entry : entries) {
                List<ClusterNode> nodes;
                try {
                    KeyCacheObject key = entry.getKey();
                    assert key != null;
                    if (initPda) {
                        if (cacheObjCtx.addDeploymentInfo())
                            jobPda = new DataStreamerPda(key.value(cacheObjCtx, false), entry.getValue() != null ? entry.getValue().value(cacheObjCtx, false) : null, rcvr);
                        else if (rcvr != null)
                            jobPda = new DataStreamerPda(rcvr);
                        initPda = false;
                    }
                    nodes = nodes(key, topVer, cctx);
                } catch (IgniteCheckedException e) {
                    resFut.onDone(e);
                    return;
                }
                if (F.isEmpty(nodes)) {
                    resFut.onDone(new ClusterTopologyException("Failed to map key to node " + "(no nodes with cache found in topology) [infos=" + entries.size() + ", cacheName=" + cacheName + ']'));
                    return;
                }
                for (ClusterNode node : nodes) {
                    Collection<DataStreamerEntry> col = mappings.get(node);
                    if (col == null)
                        mappings.put(node, col = new ArrayList<>());
                    col.add(entry);
                }
            }
            for (final Map.Entry<ClusterNode, Collection<DataStreamerEntry>> e : mappings.entrySet()) {
                final UUID nodeId = e.getKey().id();
                Buffer buf = bufMappings.get(nodeId);
                if (buf == null) {
                    Buffer old = bufMappings.putIfAbsent(nodeId, buf = new Buffer(e.getKey()));
                    if (old != null)
                        buf = old;
                }
                final Collection<DataStreamerEntry> entriesForNode = e.getValue();
                IgniteInClosure<IgniteInternalFuture<?>> lsnr = new IgniteInClosure<IgniteInternalFuture<?>>() {

                    @Override
                    public void apply(IgniteInternalFuture<?> t) {
                        try {
                            t.get();
                            if (activeKeys != null) {
                                for (DataStreamerEntry e : entriesForNode) activeKeys.remove(new KeyCacheObjectWrapper(e.getKey()));
                                if (activeKeys.isEmpty())
                                    resFut.onDone();
                            } else {
                                assert entriesForNode.size() == 1;
                                // That has been a single key,
                                // so complete result future right away.
                                resFut.onDone();
                            }
                        } catch (IgniteClientDisconnectedCheckedException e1) {
                            if (log.isDebugEnabled())
                                log.debug("Future finished with disconnect error [nodeId=" + nodeId + ", err=" + e1 + ']');
                            resFut.onDone(e1);
                        } catch (IgniteCheckedException e1) {
                            if (log.isDebugEnabled())
                                log.debug("Future finished with error [nodeId=" + nodeId + ", err=" + e1 + ']');
                            if (cancelled) {
                                resFut.onDone(new IgniteCheckedException("Data streamer has been cancelled: " + DataStreamerImpl.this, e1));
                            } else if (remaps + 1 > maxRemapCnt) {
                                resFut.onDone(new IgniteCheckedException("Failed to finish operation (too many remaps): " + remaps, e1));
                            } else {
                                try {
                                    remapSem.acquire();
                                    final Runnable r = new Runnable() {

                                        @Override
                                        public void run() {
                                            try {
                                                if (cancelled)
                                                    throw new IllegalStateException("DataStreamer closed.");
                                                load0(entriesForNode, resFut, activeKeys, remaps + 1);
                                            } catch (Throwable ex) {
                                                resFut.onDone(new IgniteCheckedException("DataStreamer remapping failed. ", ex));
                                            } finally {
                                                remapSem.release();
                                            }
                                        }
                                    };
                                    dataToRemap.add(r);
                                    if (!remapOwning.get() && remapOwning.compareAndSet(false, true)) {
                                        ctx.closure().callLocalSafe(new GPC<Boolean>() {

                                            @Override
                                            public Boolean call() {
                                                boolean locked = true;
                                                while (locked || !dataToRemap.isEmpty()) {
                                                    if (!locked && !remapOwning.compareAndSet(false, true))
                                                        return false;
                                                    try {
                                                        Runnable r = dataToRemap.poll();
                                                        if (r != null)
                                                            r.run();
                                                    } finally {
                                                        if (!dataToRemap.isEmpty())
                                                            locked = true;
                                                        else {
                                                            remapOwning.set(false);
                                                            locked = false;
                                                        }
                                                    }
                                                }
                                                return true;
                                            }
                                        }, true);
                                    }
                                } catch (InterruptedException e2) {
                                    resFut.onDone(e2);
                                }
                            }
                        }
                    }
                };
                final GridFutureAdapter<?> f;
                try {
                    f = buf.update(entriesForNode, topVer, lsnr, remap);
                } catch (IgniteInterruptedCheckedException e1) {
                    resFut.onDone(e1);
                    return;
                }
                if (ctx.discovery().node(nodeId) == null) {
                    if (bufMappings.remove(nodeId, buf)) {
                        final Buffer buf0 = buf;
                        waitAffinityAndRun(new Runnable() {

                            @Override
                            public void run() {
                                buf0.onNodeLeft();
                                if (f != null)
                                    f.onDone(new ClusterTopologyCheckedException("Failed to wait for request completion " + "(node has left): " + nodeId));
                            }
                        }, ctx.discovery().topologyVersion(), false);
                    }
                }
            }
        } finally {
            if (gate != null)
                gate.leave();
        }
    } catch (Exception ex) {
        resFut.onDone(new IgniteCheckedException("DataStreamer data loading failed.", ex));
    }
}

Example 2 with ClusterTopologyException

use of org.apache.ignite.cluster.ClusterTopologyException in project ignite by apache.

the class GridTaskWorker method sendRequest.

/**
     * @param res Job result.
     */
private void sendRequest(ComputeJobResult res) {
    assert res != null;
    GridJobExecuteRequest req = null;
    ClusterNode node = res.getNode();
    try {
        ClusterNode curNode = ctx.discovery().node(node.id());
        // thrown in case of send failure.
        if (curNode == null) {
            U.warn(log, "Failed to send job request because remote node left grid (if fail-over is enabled, " + "will attempt fail-over to another node) [node=" + node + ", taskName=" + ses.getTaskName() + ", taskSesId=" + ses.getId() + ", jobSesId=" + res.getJobContext().getJobId() + ']');
            ctx.resource().invokeAnnotated(dep, res.getJob(), ComputeJobAfterSend.class);
            GridJobExecuteResponse fakeRes = new GridJobExecuteResponse(node.id(), ses.getId(), res.getJobContext().getJobId(), null, null, null, null, null, null, false, null);
            fakeRes.setFakeException(new ClusterTopologyException("Failed to send job due to node failure: " + node));
            onResponse(fakeRes);
        } else {
            long timeout = ses.getEndTime() == Long.MAX_VALUE ? Long.MAX_VALUE : ses.getEndTime() - U.currentTimeMillis();
            if (timeout > 0) {
                boolean loc = node.id().equals(ctx.discovery().localNode().id()) && !ctx.config().isMarshalLocalJobs();
                Map<Object, Object> sesAttrs = ses.isFullSupport() ? ses.getAttributes() : null;
                Map<? extends Serializable, ? extends Serializable> jobAttrs = (Map<? extends Serializable, ? extends Serializable>) res.getJobContext().getAttributes();
                boolean forceLocDep = internal || !ctx.deploy().enabled();
                req = new GridJobExecuteRequest(ses.getId(), res.getJobContext().getJobId(), ses.getTaskName(), ses.getUserVersion(), ses.getTaskClassName(), loc ? null : U.marshal(marsh, res.getJob()), loc ? res.getJob() : null, ses.getStartTime(), timeout, ses.getTopology(), loc ? null : U.marshal(marsh, ses.getJobSiblings()), loc ? ses.getJobSiblings() : null, loc ? null : U.marshal(marsh, sesAttrs), loc ? sesAttrs : null, loc ? null : U.marshal(marsh, jobAttrs), loc ? jobAttrs : null, ses.getCheckpointSpi(), dep.classLoaderId(), dep.deployMode(), continuous, dep.participants(), forceLocDep, ses.isFullSupport(), internal, subjId, affCacheIds, affPartId, mapTopVer, ses.executorName());
                if (loc)
                    ctx.job().processJobExecuteRequest(ctx.discovery().localNode(), req);
                else {
                    byte plc;
                    if (internal)
                        plc = MANAGEMENT_POOL;
                    else {
                        Byte ctxPlc = getThreadContext(TC_IO_POLICY);
                        if (ctxPlc != null)
                            plc = ctxPlc;
                        else
                            plc = PUBLIC_POOL;
                    }
                    // Send job execution request.
                    ctx.io().sendToGridTopic(node, TOPIC_JOB, req, plc);
                    if (log.isDebugEnabled())
                        log.debug("Sent job request [req=" + req + ", node=" + node + ']');
                }
                if (!loc)
                    ctx.resource().invokeAnnotated(dep, res.getJob(), ComputeJobAfterSend.class);
            } else
                U.warn(log, "Job timed out prior to sending job execution request: " + res.getJob());
        }
    } catch (IgniteCheckedException e) {
        IgniteException fakeErr = null;
        try {
            boolean deadNode = isDeadNode(res.getNode().id());
            // Avoid stack trace if node has left grid.
            if (deadNode) {
                U.warn(log, "Failed to send job request because remote node left grid (if failover is enabled, " + "will attempt fail-over to another node) [node=" + node + ", taskName=" + ses.getTaskName() + ", taskSesId=" + ses.getId() + ", jobSesId=" + res.getJobContext().getJobId() + ']');
                fakeErr = new ClusterTopologyException("Failed to send job due to node failure: " + node, e);
            } else
                U.error(log, "Failed to send job request: " + req, e);
        } catch (IgniteClientDisconnectedCheckedException e0) {
            if (log.isDebugEnabled())
                log.debug("Failed to send job request, client disconnected [node=" + node + ", taskName=" + ses.getTaskName() + ", taskSesId=" + ses.getId() + ", jobSesId=" + res.getJobContext().getJobId() + ']');
            fakeErr = U.convertException(e0);
        }
        GridJobExecuteResponse fakeRes = new GridJobExecuteResponse(node.id(), ses.getId(), res.getJobContext().getJobId(), null, null, null, null, null, null, false, null);
        if (fakeErr == null)
            fakeErr = U.convertException(e);
        fakeRes.setFakeException(fakeErr);
        onResponse(fakeRes);
    }
}

Example 3 with ClusterTopologyException

use of org.apache.ignite.cluster.ClusterTopologyException in project ignite by apache.

the class CacheContinuousQueryFailoverAbstractSelfTest method checkBackupQueue.

/**
     * @param backups Number of backups.
     * @param updateFromClient If {@code true} executes cache update from client node.
     * @throws Exception If failed.
     */
private void checkBackupQueue(int backups, boolean updateFromClient) throws Exception {
    this.backups = atomicityMode() == CacheAtomicityMode.ATOMIC ? backups : backups < 2 ? 2 : backups;
    final int SRV_NODES = 4;
    startGridsMultiThreaded(SRV_NODES);
    client = true;
    Ignite qryClient = startGrid(SRV_NODES);
    client = false;
    IgniteCache<Object, Object> qryClientCache = qryClient.cache(DEFAULT_CACHE_NAME);
    Affinity<Object> aff = qryClient.affinity(DEFAULT_CACHE_NAME);
    CacheEventListener1 lsnr = asyncCallback() ? new CacheEventAsyncListener1(false) : new CacheEventListener1(false);
    ContinuousQuery<Object, Object> qry = new ContinuousQuery<>();
    qry.setLocalListener(lsnr);
    QueryCursor<?> cur = qryClientCache.query(qry);
    int PARTS = 10;
    Map<Object, T2<Object, Object>> updates = new HashMap<>();
    List<T3<Object, Object, Object>> expEvts = new ArrayList<>();
    for (int i = 0; i < (atomicityMode() == CacheAtomicityMode.ATOMIC ? SRV_NODES - 1 : SRV_NODES - 2); i++) {
        log.info("Stop iteration: " + i);
        TestCommunicationSpi spi = (TestCommunicationSpi) ignite(i).configuration().getCommunicationSpi();
        Ignite ignite = ignite(i);
        IgniteCache<Object, Object> cache = ignite.cache(DEFAULT_CACHE_NAME);
        List<Integer> keys = testKeys(cache, PARTS);
        CountDownLatch latch = new CountDownLatch(keys.size());
        lsnr.latch = latch;
        boolean first = true;
        for (Integer key : keys) {
            log.info("Put [node=" + ignite.name() + ", key=" + key + ", part=" + aff.partition(key) + ']');
            T2<Object, Object> t = updates.get(key);
            if (updateFromClient) {
                if (atomicityMode() == CacheAtomicityMode.TRANSACTIONAL) {
                    try (Transaction tx = qryClient.transactions().txStart()) {
                        qryClientCache.put(key, key);
                        tx.commit();
                    } catch (CacheException | ClusterTopologyException ignored) {
                        log.warning("Failed put. [Key=" + key + ", val=" + key + "]");
                        continue;
                    }
                } else
                    qryClientCache.put(key, key);
            } else {
                if (atomicityMode() == CacheAtomicityMode.TRANSACTIONAL) {
                    try (Transaction tx = ignite.transactions().txStart()) {
                        cache.put(key, key);
                        tx.commit();
                    } catch (CacheException | ClusterTopologyException ignored) {
                        log.warning("Failed put. [Key=" + key + ", val=" + key + "]");
                        continue;
                    }
                } else
                    cache.put(key, key);
            }
            if (t == null) {
                updates.put(key, new T2<>((Object) key, null));
                expEvts.add(new T3<>((Object) key, (Object) key, null));
            } else {
                updates.put(key, new T2<>((Object) key, (Object) key));
                expEvts.add(new T3<>((Object) key, (Object) key, (Object) key));
            }
            if (first) {
                spi.skipMsg = true;
                first = false;
            }
        }
        stopGrid(i);
        if (!latch.await(5, SECONDS)) {
            Set<Integer> keys0 = new HashSet<>(keys);
            keys0.removeAll(lsnr.keys);
            log.info("Missed events for keys: " + keys0);
            fail("Failed to wait for notifications [exp=" + keys.size() + ", left=" + lsnr.latch.getCount() + ']');
        }
        checkEvents(expEvts, lsnr);
    }
    for (int i = 0; i < (atomicityMode() == CacheAtomicityMode.ATOMIC ? SRV_NODES - 1 : SRV_NODES - 2); i++) {
        log.info("Start iteration: " + i);
        Ignite ignite = startGrid(i);
        IgniteCache<Object, Object> cache = ignite.cache(DEFAULT_CACHE_NAME);
        List<Integer> keys = testKeys(cache, PARTS);
        CountDownLatch latch = new CountDownLatch(keys.size());
        lsnr.latch = latch;
        for (Integer key : keys) {
            log.info("Put [node=" + ignite.name() + ", key=" + key + ", part=" + aff.partition(key) + ']');
            T2<Object, Object> t = updates.get(key);
            if (t == null) {
                updates.put(key, new T2<>((Object) key, null));
                expEvts.add(new T3<>((Object) key, (Object) key, null));
            } else {
                updates.put(key, new T2<>((Object) key, (Object) key));
                expEvts.add(new T3<>((Object) key, (Object) key, (Object) key));
            }
            if (updateFromClient)
                qryClientCache.put(key, key);
            else
                cache.put(key, key);
        }
        if (!latch.await(10, SECONDS)) {
            Set<Integer> keys0 = new HashSet<>(keys);
            keys0.removeAll(lsnr.keys);
            log.info("Missed events for keys: " + keys0);
            fail("Failed to wait for notifications [exp=" + keys.size() + ", left=" + lsnr.latch.getCount() + ']');
        }
        checkEvents(expEvts, lsnr);
    }
    cur.close();
    assertFalse("Unexpected error during test, see log for details.", err);
}

Example 4 with ClusterTopologyException

use of org.apache.ignite.cluster.ClusterTopologyException in project ignite by apache.

the class GridFailoverTaskWithPredicateSelfTest method testJobFailedOver.

/**
     * Tests that failover happens on three-node grid when the Task is applicable for the first node
     * and fails on it, but is also applicable on another node.
     *
     * @throws Exception If failed.
     */
public void testJobFailedOver() throws Exception {
    failed.set(false);
    routed.set(false);
    try {
        Ignite ignite1 = startGrid(NODE1);
        Ignite ignite2 = startGrid(NODE2);
        Ignite ignite3 = startGrid(NODE3);
        assert ignite1 != null;
        assert ignite2 != null;
        assert ignite3 != null;
        Integer res = (Integer) compute(ignite1.cluster().forPredicate(p)).withTimeout(10000).execute(JobFailTask.class.getName(), "1");
        assert res == 1;
    } catch (ClusterTopologyException ignored) {
        failed.set(true);
    } finally {
        assertFalse(failed.get());
        assertTrue(routed.get());
        stopGrid(NODE1);
        stopGrid(NODE2);
        stopGrid(NODE3);
    }
}

Example 5 with ClusterTopologyException

use of org.apache.ignite.cluster.ClusterTopologyException in project ignite by apache.

the class GridFailoverTaskWithPredicateSelfTest method testJobNotFailedOver.

/**
     * Tests that failover doesn't happen on two-node grid when the Task is applicable only for the first node
     * and fails on it.
     *
     * @throws Exception If failed.
     */
public void testJobNotFailedOver() throws Exception {
    failed.set(false);
    routed.set(false);
    try {
        Ignite ignite1 = startGrid(NODE1);
        Ignite ignite2 = startGrid(NODE2);
        assert ignite1 != null;
        assert ignite2 != null;
        compute(ignite1.cluster().forPredicate(p)).withTimeout(10000).execute(JobFailTask.class.getName(), "1");
    } catch (ClusterTopologyException ignored) {
        failed.set(true);
    } finally {
        assertTrue(failed.get());
        assertFalse(routed.get());
        stopGrid(NODE1);
        stopGrid(NODE2);
    }
}