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

Example 41 with ZooKeeper

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

the class SnapshotSaveAPI method logParticipatingHostCount.

/**
     * Once participating host count is set, SnapshotCompletionMonitor can check this ZK node to
     * determine whether the snapshot has finished or not.
     *
     * This should only be called when all participants have responded. It is possible that some
     * hosts finish taking snapshot before the coordinator logs the participating host count. In
     * this case, the host count would have been decremented multiple times already. To make sure
     * finished hosts are logged correctly, this method adds participating host count + 1 to the
     * current host count.
     *
     * @param txnId The snapshot txnId
     * @param participantCount The number of hosts participating in this snapshot
     */
public static void logParticipatingHostCount(long txnId, int participantCount) {
    ZooKeeper zk = VoltDB.instance().getHostMessenger().getZK();
    final String snapshotPath = VoltZK.completed_snapshots + "/" + txnId;
    boolean success = false;
    while (!success) {
        Stat stat = new Stat();
        byte[] data = null;
        try {
            data = zk.getData(snapshotPath, false, stat);
        } catch (KeeperException e) {
            if (e.code() == KeeperException.Code.NONODE) {
                // If snapshot creation failed for some reason, the node won't exist. ignore
                return;
            }
            VoltDB.crashLocalVoltDB("Failed to get snapshot completion node", true, e);
        } catch (InterruptedException e) {
            VoltDB.crashLocalVoltDB("Interrupted getting snapshot completion node", true, e);
        }
        if (data == null) {
            VoltDB.crashLocalVoltDB("Data should not be null if the node exists", false, null);
        }
        try {
            JSONObject jsonObj = new JSONObject(new String(data, Charsets.UTF_8));
            if (jsonObj.getLong("txnId") != txnId) {
                VoltDB.crashLocalVoltDB("TxnId should match", false, null);
            }
            int hostCount = jsonObj.getInt("hostCount");
            // +1 because hostCount was initialized to -1
            jsonObj.put("hostCount", hostCount + participantCount + 1);
            zk.setData(snapshotPath, jsonObj.toString(4).getBytes(Charsets.UTF_8), stat.getVersion());
        } catch (KeeperException.BadVersionException e) {
            continue;
        } catch (Exception e) {
            VoltDB.crashLocalVoltDB("This ZK call should never fail", true, e);
        }
        success = true;
    }
}

Example 42 with ZooKeeper

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

the class RealVoltDB method recoveryComplete.

@Override
public synchronized void recoveryComplete(String requestId) {
    assert (m_rejoinDataPending == false);
    if (m_rejoining) {
        if (m_rejoinTruncationReqId.compareTo(requestId) <= 0) {
            String actionName = m_joining ? "join" : "rejoin";
            // remove the rejoin blocker
            CoreZK.removeRejoinNodeIndicatorForHost(m_messenger.getZK(), m_myHostId);
            consoleLog.info(String.format("Node %s completed", actionName));
            m_rejoinTruncationReqId = null;
            m_rejoining = false;
        } else {
            // don't flip the m_rejoining state, all truncation snapshot completions will call back to here.
            try {
                final ZooKeeper zk = m_messenger.getZK();
                String requestNode = zk.create(VoltZK.request_truncation_snapshot_node, null, Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT_SEQUENTIAL);
                if (m_rejoinTruncationReqId == null) {
                    m_rejoinTruncationReqId = requestNode;
                }
            } catch (Exception e) {
                VoltDB.crashLocalVoltDB("Unable to retry post-rejoin truncation snapshot request.", true, e);
            }
        }
    }
}

Example 43 with ZooKeeper

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

the class SnapshotSiteProcessor method logSnapshotCompleteToZK.

private static void logSnapshotCompleteToZK(long txnId, boolean snapshotSuccess, ExtensibleSnapshotDigestData extraSnapshotData) {
    ZooKeeper zk = VoltDB.instance().getHostMessenger().getZK();
    // Timeout after 10 minutes
    final long endTime = System.currentTimeMillis() + TimeUnit.MINUTES.toMillis(10);
    final String snapshotPath = VoltZK.completed_snapshots + "/" + txnId;
    boolean success = false;
    while (!success) {
        if (System.currentTimeMillis() > endTime) {
            VoltDB.crashLocalVoltDB("Timed out logging snapshot completion to ZK");
        }
        Stat stat = new Stat();
        byte[] data = null;
        try {
            data = zk.getData(snapshotPath, false, stat);
        } catch (NoNodeException e) {
            // if the node doesn't exist yet, retry
            continue;
        } catch (Exception e) {
            VoltDB.crashLocalVoltDB("This ZK get should never fail", true, e);
        }
        if (data == null) {
            VoltDB.crashLocalVoltDB("Data should not be null if the node exists", false, null);
        }
        try {
            JSONObject jsonObj = new JSONObject(new String(data, "UTF-8"));
            if (jsonObj.getLong("txnId") != txnId) {
                VoltDB.crashLocalVoltDB("TxnId should match", false, null);
            }
            int remainingHosts = jsonObj.getInt("hostCount") - 1;
            jsonObj.put("hostCount", remainingHosts);
            jsonObj.put("didSucceed", snapshotSuccess);
            if (!snapshotSuccess) {
                jsonObj.put("isTruncation", false);
            }
            extraSnapshotData.mergeToZooKeeper(jsonObj, SNAP_LOG);
            byte[] zkData = jsonObj.toString().getBytes("UTF-8");
            if (zkData.length > 5000000) {
                SNAP_LOG.warn("ZooKeeper node for snapshot digest unexpectedly large: " + zkData.length);
            }
            zk.setData(snapshotPath, zkData, stat.getVersion());
        } catch (KeeperException.BadVersionException e) {
            continue;
        } catch (Exception e) {
            VoltDB.crashLocalVoltDB("This ZK call should never fail", true, e);
        }
        success = true;
    }
    /*
         * If we are running without command logging there will be no consumer for
         * the completed snapshot messages. Consume them here to bound space usage in ZK.
         */
    try {
        TreeSet<String> snapshots = new TreeSet<String>(zk.getChildren(VoltZK.completed_snapshots, false));
        while (snapshots.size() > 30) {
            try {
                zk.delete(VoltZK.completed_snapshots + "/" + snapshots.first(), -1);
            } catch (NoNodeException e) {
            } catch (Exception e) {
                VoltDB.crashLocalVoltDB("Deleting a snapshot completion record from ZK should only fail with NoNodeException", true, e);
            }
            snapshots.remove(snapshots.first());
        }
    } catch (Exception e) {
        VoltDB.crashLocalVoltDB("Retrieving list of completed snapshots from ZK should never fail", true, e);
    }
}

Example 44 with ZooKeeper

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

the class MockVoltDB method validateStartAction.

public void validateStartAction() {
    try {
        ZooKeeper zk = m_hostMessenger.getZK();
        boolean initCompleted = zk.exists(VoltZK.init_completed, false) != null;
        List<String> children = zk.getChildren(VoltZK.start_action, new StartActionWatcher(), null);
        if (!children.isEmpty()) {
            for (String child : children) {
                byte[] data = zk.getData(VoltZK.start_action + "/" + child, false, null);
                if (data == null) {
                    VoltDB.crashLocalVoltDB("Couldn't find " + VoltZK.start_action + "/" + child);
                }
                String startAction = new String(data);
                if ((startAction.equals(StartAction.JOIN.toString()) || startAction.equals(StartAction.REJOIN.toString()) || startAction.equals(StartAction.LIVE_REJOIN.toString())) && !initCompleted) {
                    int nodeId = VoltZK.getHostIDFromChildName(child);
                    if (nodeId == m_hostMessenger.getHostId()) {
                        VoltDB.crashLocalVoltDB("This node was started with start action " + startAction + " during cluster creation. All nodes should be started with matching " + "create or recover actions when bring up a cluster. Join and Rejoin " + "are for adding nodes to an already running cluster.");
                    } else {
                        logger.warn("Node " + nodeId + " tried to " + startAction + " cluster but it is not allowed during cluster creation. " + "All nodes should be started with matching create or recover actions when bring up a cluster. " + "Join and rejoin are for adding nodes to an already running cluster.");
                    }
                }
            }
        }
    } catch (KeeperException e) {
        logger.error("Failed to validate the start actions:" + e.getMessage());
    } catch (InterruptedException e) {
        VoltDB.crashLocalVoltDB("Interrupted during start action validation:" + e.getMessage(), true, e);
    }
}

Example 45 with ZooKeeper

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

the class TestMapCache method testDeleteChild.

@Test
public void testDeleteChild() throws Exception {
    ZooKeeper zk = getClient(0);
    configure("/cache02", zk);
    MapCache dut = new MapCache(zk, "/cache02");
    dut.start(true);
    Map<String, JSONObject> cache = dut.pointInTimeCache();
    assertEquals("3 items cached.", 3, cache.size());
    zk.delete("/cache02/bb", -1);
    while (true) {
        cache = dut.pointInTimeCache();
        if (cache.size() == 3) {
            Thread.sleep(1);
        } else {
            break;
        }
    }
    assertEquals("Item removed", 2, cache.size());
    assertEquals(null, cache.get("/cache02/bb"));
    assertEquals("aaval", cache.get("/cache02/aa").get("key"));
    assertEquals("ccval", cache.get("/cache02/cc").get("key"));
    dut.shutdown();
    zk.close();
}