Examples with ZooKeeper org.apache.zookeeper_voltpatches.ZooKeeper used on opensource projects

Example 1 with ZooKeeper

use of org.apache.zookeeper_voltpatches.ZooKeeper in project voltdb by VoltDB.

the class InvocationDispatcher method takeShutdownSaveSnapshot.

private final ClientResponseImpl takeShutdownSaveSnapshot(final StoredProcedureInvocation task, final InvocationClientHandler handler, final Connection ccxn, final AuthUser user, OverrideCheck bypass) {
    // shutdown save snapshot is available for Pro edition only
    if (!MiscUtils.isPro()) {
        task.setParams();
        return dispatch(task, handler, ccxn, user, bypass, false);
    }
    Object p0 = task.getParams().getParam(0);
    final long zkTxnId;
    if (p0 instanceof Long) {
        zkTxnId = ((Long) p0).longValue();
    } else if (p0 instanceof String) {
        try {
            zkTxnId = Long.parseLong((String) p0);
        } catch (NumberFormatException e) {
            return gracefulFailureResponse("Incorrect argument type", task.clientHandle);
        }
    } else {
        return gracefulFailureResponse("Incorrect argument type", task.clientHandle);
    }
    VoltDBInterface voltdb = VoltDB.instance();
    if (!voltdb.isPreparingShuttingdown()) {
        log.warn("Ignoring shutdown save snapshot request as VoltDB is not shutting down");
        return unexpectedFailureResponse("Ignoring shutdown save snapshot request as VoltDB is not shutting down", task.clientHandle);
    }
    final ZooKeeper zk = voltdb.getHostMessenger().getZK();
    // network threads are blocked from making zookeeper calls
    Future<Long> fut = voltdb.getSES(true).submit(new Callable<Long>() {

        @Override
        public Long call() {
            try {
                Stat stat = zk.exists(VoltZK.operationMode, false);
                if (stat == null) {
                    VoltDB.crashLocalVoltDB("cluster operation mode zookeeper node does not exist");
                    return Long.MIN_VALUE;
                }
                return stat.getMzxid();
            } catch (KeeperException | InterruptedException e) {
                VoltDB.crashLocalVoltDB("Failed to stat the cluster operation zookeeper node", true, e);
                return Long.MIN_VALUE;
            }
        }
    });
    try {
        if (fut.get().longValue() != zkTxnId) {
            return unexpectedFailureResponse("Internal error: cannot write a startup snapshot because the " + "current system state is not consistent with an orderly shutdown. " + "Please try \"voltadmin shutdown --save\" again.", task.clientHandle);
        }
    } catch (InterruptedException | ExecutionException e1) {
        VoltDB.crashLocalVoltDB("Failed to stat the cluster operation zookeeper node", true, e1);
        return null;
    }
    NodeSettings paths = m_catalogContext.get().getNodeSettings();
    String data;
    try {
        data = new JSONStringer().object().keySymbolValuePair(SnapshotUtil.JSON_TERMINUS, zkTxnId).endObject().toString();
    } catch (JSONException e) {
        VoltDB.crashLocalVoltDB("Failed to create startup snapshot save command", true, e);
        return null;
    }
    log.info("Saving startup snapshot");
    consoleLog.info("Taking snapshot to save database contents");
    final SimpleClientResponseAdapter alternateAdapter = new SimpleClientResponseAdapter(ClientInterface.SHUTDONW_SAVE_CID, "Blocking Startup Snapshot Save");
    final InvocationClientHandler alternateHandler = new InvocationClientHandler() {

        @Override
        public boolean isAdmin() {
            return handler.isAdmin();
        }

        @Override
        public long connectionId() {
            return ClientInterface.SHUTDONW_SAVE_CID;
        }
    };
    final long sourceHandle = task.clientHandle;
    task.setClientHandle(alternateAdapter.registerCallback(SimpleClientResponseAdapter.NULL_CALLBACK));
    SnapshotUtil.SnapshotResponseHandler savCallback = new SnapshotUtil.SnapshotResponseHandler() {

        @Override
        public void handleResponse(ClientResponse r) {
            if (r == null) {
                String msg = "Snapshot save failed. The database is paused and the shutdown has been cancelled";
                transmitResponseMessage(gracefulFailureResponse(msg, sourceHandle), ccxn, sourceHandle);
            }
            if (r.getStatus() != ClientResponse.SUCCESS) {
                String msg = "Snapshot save failed: " + r.getStatusString() + ". The database is paused and the shutdown has been cancelled";
                ClientResponseImpl resp = new ClientResponseImpl(ClientResponse.GRACEFUL_FAILURE, r.getResults(), msg, sourceHandle);
                transmitResponseMessage(resp, ccxn, sourceHandle);
            }
            consoleLog.info("Snapshot taken successfully");
            task.setParams();
            dispatch(task, alternateHandler, alternateAdapter, user, bypass, false);
        }
    };
    // network threads are blocked from making zookeeper calls
    final byte[] guardContent = data.getBytes(StandardCharsets.UTF_8);
    Future<Boolean> guardFuture = voltdb.getSES(true).submit(new Callable<Boolean>() {

        @Override
        public Boolean call() throws Exception {
            try {
                ZKUtil.asyncMkdirs(zk, VoltZK.shutdown_save_guard, guardContent).get();
            } catch (NodeExistsException itIsOk) {
                return false;
            } catch (InterruptedException | KeeperException e) {
                VoltDB.crashLocalVoltDB("Failed to create shutdown save guard zookeeper node", true, e);
                return false;
            }
            return true;
        }
    });
    boolean created;
    try {
        created = guardFuture.get().booleanValue();
    } catch (InterruptedException | ExecutionException e) {
        VoltDB.crashLocalVoltDB("Failed to create shutdown save guard zookeeper node", true, e);
        return null;
    }
    if (!created) {
        return unexpectedFailureResponse("Internal error: detected concurrent invocations of \"voltadmin shutdown --save\"", task.clientHandle);
    }
    voltdb.getClientInterface().bindAdapter(alternateAdapter, null);
    SnapshotUtil.requestSnapshot(sourceHandle, paths.resolve(paths.getSnapshoth()).toPath().toUri().toString(), SnapshotUtil.getShutdownSaveNonce(zkTxnId), true, SnapshotFormat.NATIVE, SnapshotPathType.SNAP_AUTO, data, savCallback, true);
    return null;
}

Example 2 with ZooKeeper

use of org.apache.zookeeper_voltpatches.ZooKeeper in project voltdb by VoltDB.

the class VoltZK method updateClusterMetadata.

public static void updateClusterMetadata(Map<Integer, String> clusterMetadata) throws Exception {
    ZooKeeper zk = VoltDB.instance().getHostMessenger().getZK();
    List<String> metadataNodes = zk.getChildren(VoltZK.cluster_metadata, false);
    Set<Integer> hostIds = new HashSet<Integer>();
    for (String hostId : metadataNodes) {
        hostIds.add(Integer.valueOf(hostId));
    }
    /*
         * Remove anything that is no longer part of the cluster
         */
    Set<Integer> keySetCopy = new HashSet<Integer>(clusterMetadata.keySet());
    keySetCopy.removeAll(hostIds);
    for (Integer failedHostId : keySetCopy) {
        clusterMetadata.remove(failedHostId);
    }
    /*
         * Add anything that is new
         */
    Set<Integer> hostIdsCopy = new HashSet<Integer>(hostIds);
    hostIdsCopy.removeAll(clusterMetadata.keySet());
    List<Pair<Integer, ZKUtil.ByteArrayCallback>> callbacks = new ArrayList<Pair<Integer, ZKUtil.ByteArrayCallback>>();
    for (Integer hostId : hostIdsCopy) {
        ZKUtil.ByteArrayCallback cb = new ZKUtil.ByteArrayCallback();
        callbacks.add(Pair.of(hostId, cb));
        zk.getData(VoltZK.cluster_metadata + "/" + hostId, false, cb, null);
    }
    for (Pair<Integer, ZKUtil.ByteArrayCallback> p : callbacks) {
        Integer hostId = p.getFirst();
        ZKUtil.ByteArrayCallback cb = p.getSecond();
        try {
            clusterMetadata.put(hostId, new String(cb.getData(), "UTF-8"));
        } catch (KeeperException.NoNodeException e) {
        }
    }
}

Example 3 with ZooKeeper

use of org.apache.zookeeper_voltpatches.ZooKeeper in project voltdb by VoltDB.

the class UpdateSettings method run.

public VoltTable[] run(SystemProcedureExecutionContext ctx, byte[] settingsBytes) {
    ZooKeeper zk = getHostMessenger().getZK();
    Stat stat = null;
    try {
        stat = zk.exists(VoltZK.cluster_settings, false);
    } catch (KeeperException | InterruptedException e) {
        String msg = "Failed to stat cluster settings zookeeper node";
        log.error(msg, e);
        throw new VoltAbortException(msg);
    }
    final int version = stat.getVersion();
    executeSysProcPlanFragments(createBarrierFragment(settingsBytes, version), DEP_updateSettingsBarrierAggregate);
    return executeSysProcPlanFragments(createUpdateFragment(settingsBytes, version), DEP_updateSettingsAggregate);
}

Example 4 with ZooKeeper

use of org.apache.zookeeper_voltpatches.ZooKeeper in project voltdb by VoltDB.

the class UpdateSettings method executePlanFragment.

@Override
public DependencyPair executePlanFragment(Map<Integer, List<VoltTable>> dependencies, long fragmentId, ParameterSet params, SystemProcedureExecutionContext context) {
    if (fragmentId == SysProcFragmentId.PF_updateSettingsBarrier) {
        DependencyPair success = new DependencyPair.TableDependencyPair(DEP_updateSettingsBarrier, new VoltTable(new ColumnInfo[] { new ColumnInfo("UNUSED", VoltType.BIGINT) }));
        if (log.isInfoEnabled()) {
            log.info("Site " + CoreUtils.hsIdToString(m_site.getCorrespondingSiteId()) + " reached settings update barrier.");
        }
        return success;
    } else if (fragmentId == SysProcFragmentId.PF_updateSettingsBarrierAggregate) {
        Object[] paramarr = params.toArray();
        byte[] settingsBytes = (byte[]) paramarr[0];
        int version = ((Integer) paramarr[1]).intValue();
        ZooKeeper zk = getHostMessenger().getZK();
        Stat stat = null;
        try {
            stat = zk.setData(VoltZK.cluster_settings, settingsBytes, version);
        } catch (KeeperException | InterruptedException e) {
            String msg = "Failed to update cluster settings";
            log.error(msg, e);
            throw new SettingsException(msg, e);
        }
        log.info("Saved new cluster settings state");
        return new DependencyPair.TableDependencyPair(DEP_updateSettingsBarrierAggregate, getVersionResponse(stat.getVersion()));
    } else if (fragmentId == SysProcFragmentId.PF_updateSettings) {
        Object[] paramarr = params.toArray();
        byte[] settingsBytes = (byte[]) paramarr[0];
        int version = ((Integer) paramarr[1]).intValue();
        ClusterSettings settings = ClusterSettings.create(settingsBytes);
        Pair<CatalogContext, CatalogSpecificPlanner> ctgdef = getVoltDB().settingsUpdate(settings, version);
        context.updateSettings(ctgdef.getFirst(), ctgdef.getSecond());
        VoltTable result = new VoltTable(VoltSystemProcedure.STATUS_SCHEMA);
        result.addRow(VoltSystemProcedure.STATUS_OK);
        return new DependencyPair.TableDependencyPair(DEP_updateSettings, result);
    } else if (fragmentId == SysProcFragmentId.PF_updateSettingsAggregate) {
        VoltTable result = VoltTableUtil.unionTables(dependencies.get(DEP_updateSettings));
        return new DependencyPair.TableDependencyPair(DEP_updateSettingsAggregate, result);
    } else {
        VoltDB.crashLocalVoltDB("Received unrecognized plan fragment id " + fragmentId + " in UpdateSettings", false, null);
    }
    throw new RuntimeException("Should not reach this code");
}

Example 5 with ZooKeeper

use of org.apache.zookeeper_voltpatches.ZooKeeper in project voltdb by VoltDB.

the class MeshProber method considerMeshPlea.

@Override
public JoinAcceptor.PleaDecision considerMeshPlea(ZooKeeper zk, int hostId, JSONObject jo) {
    checkArgument(zk != null, "zookeeper is null");
    checkArgument(jo != null, "json object is null");
    if (!HostCriteria.hasCriteria(jo)) {
        return new JoinAcceptor.PleaDecision(String.format("Joining node version %s is incompatible with this node verion %s", jo.optString(SocketJoiner.VERSION_STRING, "(unknown)"), m_versionChecker.getVersionString()), false, false);
    }
    HostCriteria hc = new HostCriteria(jo);
    Map<Integer, HostCriteria> hostCriteria = m_hostCriteria.get();
    // when the cluster is forming anew)
    if (!getNodeState().operational() && !hostCriteria.values().stream().anyMatch(c -> c.getNodeState().operational())) {
        List<String> incompatibilities = asHostCriteria().listIncompatibilities(hc);
        if (!incompatibilities.isEmpty()) {
            Joiner joiner = Joiner.on("\n    ").skipNulls();
            String error = "Incompatible joining criteria:\n    " + joiner.join(incompatibilities);
            return new JoinAcceptor.PleaDecision(error, false, false);
        }
        return new JoinAcceptor.PleaDecision(null, true, false);
    } else {
        StartAction operationalStartAction = hostCriteria.values().stream().filter(c -> c.getNodeState().operational()).map(c -> c.getStartAction()).findFirst().orElse(getStartAction());
        if (operationalStartAction == StartAction.PROBE && hc.getStartAction() != StartAction.PROBE) {
            String msg = "Invalid VoltDB command. Please use init and start to join this cluster";
            return new JoinAcceptor.PleaDecision(msg, false, false);
        }
    }
    // how many hosts are already in the mesh?
    Stat stat = new Stat();
    try {
        zk.getChildren(CoreZK.hosts, false, stat);
    } catch (InterruptedException e) {
        String msg = "Interrupted while considering mesh plea";
        m_networkLog.error(msg, e);
        return new JoinAcceptor.PleaDecision(msg, false, false);
    } catch (KeeperException e) {
        EnumSet<KeeperException.Code> closing = EnumSet.of(KeeperException.Code.SESSIONEXPIRED, KeeperException.Code.CONNECTIONLOSS);
        if (closing.contains(e.code())) {
            return new JoinAcceptor.PleaDecision("Shutting down", false, false);
        } else {
            String msg = "Failed to list hosts while considering a mesh plea";
            m_networkLog.error(msg, e);
            return new JoinAcceptor.PleaDecision(msg, false, false);
        }
    }
    // connecting to already wholly formed cluster
    if (stat.getNumChildren() >= getHostCount()) {
        return new JoinAcceptor.PleaDecision(hc.isAddAllowed() ? null : "Cluster is already complete", hc.isAddAllowed(), false);
    } else if (stat.getNumChildren() < getHostCount()) {
        // check for concurrent rejoins
        final int rejoiningHost = CoreZK.createRejoinNodeIndicator(zk, hostId);
        if (rejoiningHost == -1) {
            return new JoinAcceptor.PleaDecision(null, true, false);
        } else {
            String msg = "Only one host can rejoin at a time. Host " + rejoiningHost + " is still rejoining.";
            return new JoinAcceptor.PleaDecision(msg, false, true);
        }
    }
    return new JoinAcceptor.PleaDecision(null, true, false);
}