Examples with RetryNode org.apache.solr.update.SolrCmdDistributor.RetryNode used on opensource projects

Example 1 with RetryNode

use of org.apache.solr.update.SolrCmdDistributor.RetryNode in project lucene-solr by apache.

the class SolrCmdDistributorTest method testMaxRetries.

private void testMaxRetries() throws IOException {
    final MockStreamingSolrClients streamingClients = new MockStreamingSolrClients(updateShardHandler);
    SolrCmdDistributor cmdDistrib = new SolrCmdDistributor(streamingClients, 5, 0);
    streamingClients.setExp(Exp.CONNECT_EXCEPTION);
    ArrayList<Node> nodes = new ArrayList<>();
    final HttpSolrClient solrclient1 = (HttpSolrClient) clients.get(0);
    final AtomicInteger retries = new AtomicInteger();
    ZkNodeProps nodeProps = new ZkNodeProps(ZkStateReader.BASE_URL_PROP, solrclient1.getBaseURL(), ZkStateReader.CORE_NAME_PROP, "");
    RetryNode retryNode = new RetryNode(new ZkCoreNodeProps(nodeProps), null, "collection1", "shard1") {

        @Override
        public boolean checkRetry() {
            retries.incrementAndGet();
            return true;
        }
    };
    nodes.add(retryNode);
    AddUpdateCommand cmd = new AddUpdateCommand(null);
    cmd.solrDoc = sdoc("id", id.incrementAndGet());
    ModifiableSolrParams params = new ModifiableSolrParams();
    cmdDistrib.distribAdd(cmd, nodes, params);
    cmdDistrib.finish();
    assertEquals(6, retries.get());
    assertEquals(1, cmdDistrib.getErrors().size());
}

Example 2 with RetryNode

use of org.apache.solr.update.SolrCmdDistributor.RetryNode in project lucene-solr by apache.

the class SolrCmdDistributorTest method testOneRetry.

private void testOneRetry() throws Exception {
    final HttpSolrClient solrclient = (HttpSolrClient) clients.get(0);
    long numFoundBefore = solrclient.query(new SolrQuery("*:*")).getResults().getNumFound();
    final MockStreamingSolrClients streamingClients = new MockStreamingSolrClients(updateShardHandler);
    SolrCmdDistributor cmdDistrib = new SolrCmdDistributor(streamingClients, 5, 0);
    streamingClients.setExp(Exp.CONNECT_EXCEPTION);
    ArrayList<Node> nodes = new ArrayList<>();
    ZkNodeProps nodeProps = new ZkNodeProps(ZkStateReader.BASE_URL_PROP, solrclient.getBaseURL(), ZkStateReader.CORE_NAME_PROP, "");
    final AtomicInteger retries = new AtomicInteger();
    nodeProps = new ZkNodeProps(ZkStateReader.BASE_URL_PROP, solrclient.getBaseURL(), ZkStateReader.CORE_NAME_PROP, "");
    RetryNode retryNode = new RetryNode(new ZkCoreNodeProps(nodeProps), null, "collection1", "shard1") {

        @Override
        public boolean checkRetry() {
            streamingClients.setExp(null);
            retries.incrementAndGet();
            return true;
        }
    };
    nodes.add(retryNode);
    AddUpdateCommand cmd = new AddUpdateCommand(null);
    cmd.solrDoc = sdoc("id", id.incrementAndGet());
    ModifiableSolrParams params = new ModifiableSolrParams();
    CommitUpdateCommand ccmd = new CommitUpdateCommand(null, false);
    cmdDistrib.distribAdd(cmd, nodes, params);
    cmdDistrib.distribCommit(ccmd, nodes, params);
    cmdDistrib.finish();
    assertEquals(1, retries.get());
    long numFoundAfter = solrclient.query(new SolrQuery("*:*")).getResults().getNumFound();
    // we will get java.net.ConnectException which we retry on
    assertEquals(numFoundBefore + 1, numFoundAfter);
    assertEquals(0, cmdDistrib.getErrors().size());
}

Example 3 with RetryNode

use of org.apache.solr.update.SolrCmdDistributor.RetryNode in project lucene-solr by apache.

the class DistributedUpdateProcessor method setupRequest.

private List<Node> setupRequest(String id, SolrInputDocument doc, String route) {
    List<Node> nodes = null;
    // if we are in zk mode...
    if (zkEnabled) {
        assert TestInjection.injectUpdateRandomPause();
        if ((updateCommand.getFlags() & (UpdateCommand.REPLAY | UpdateCommand.PEER_SYNC)) != 0) {
            // we actually might be the leader, but we don't want leader-logic for these types of updates anyway.
            isLeader = false;
            forwardToLeader = false;
            return nodes;
        }
        ClusterState cstate = zkController.getClusterState();
        DocCollection coll = cstate.getCollection(collection);
        Slice slice = coll.getRouter().getTargetSlice(id, doc, route, req.getParams(), coll);
        if (slice == null) {
            // No slice found.  Most strict routers will have already thrown an exception, so a null return is
            // a signal to use the slice of this core.
            // TODO: what if this core is not in the targeted collection?
            String shardId = req.getCore().getCoreDescriptor().getCloudDescriptor().getShardId();
            slice = coll.getSlice(shardId);
            if (slice == null) {
                throw new SolrException(ErrorCode.BAD_REQUEST, "No shard " + shardId + " in " + coll);
            }
        }
        DistribPhase phase = DistribPhase.parseParam(req.getParams().get(DISTRIB_UPDATE_PARAM));
        if (DistribPhase.FROMLEADER == phase && !couldIbeSubShardLeader(coll)) {
            if (req.getCore().getCoreDescriptor().getCloudDescriptor().isLeader()) {
            // locally we think we are leader but the request says it came FROMLEADER
            // that could indicate a problem, let the full logic below figure it out
            } else {
                assert TestInjection.injectFailReplicaRequests();
                // we actually might be the leader, but we don't want leader-logic for these types of updates anyway.
                isLeader = false;
                forwardToLeader = false;
                return nodes;
            }
        }
        String shardId = slice.getName();
        try {
            // Not equivalent to getLeaderProps, which does retries to find a leader.
            // Replica leader = slice.getLeader();
            Replica leaderReplica = zkController.getZkStateReader().getLeaderRetry(collection, shardId);
            isLeader = leaderReplica.getName().equals(req.getCore().getCoreDescriptor().getCloudDescriptor().getCoreNodeName());
            if (!isLeader) {
                isSubShardLeader = amISubShardLeader(coll, slice, id, doc);
                if (isSubShardLeader) {
                    String myShardId = req.getCore().getCoreDescriptor().getCloudDescriptor().getShardId();
                    slice = coll.getSlice(myShardId);
                    shardId = myShardId;
                    leaderReplica = zkController.getZkStateReader().getLeaderRetry(collection, myShardId);
                    List<ZkCoreNodeProps> myReplicas = zkController.getZkStateReader().getReplicaProps(collection, shardId, leaderReplica.getName(), null, Replica.State.DOWN);
                }
            }
            doDefensiveChecks(phase);
            // if request is coming from another collection then we want it to be sent to all replicas
            // even if its phase is FROMLEADER
            String fromCollection = updateCommand.getReq().getParams().get(DISTRIB_FROM_COLLECTION);
            if (DistribPhase.FROMLEADER == phase && !isSubShardLeader && fromCollection == null) {
                // we are coming from the leader, just go local - add no urls
                forwardToLeader = false;
            } else if (isLeader || isSubShardLeader) {
                // that means I want to forward onto my replicas...
                // so get the replicas...
                forwardToLeader = false;
                List<ZkCoreNodeProps> replicaProps = zkController.getZkStateReader().getReplicaProps(collection, shardId, leaderReplica.getName(), null, Replica.State.DOWN);
                if (replicaProps != null) {
                    if (nodes == null) {
                        nodes = new ArrayList<>(replicaProps.size());
                    }
                    // check for test param that lets us miss replicas
                    String[] skipList = req.getParams().getParams(TEST_DISTRIB_SKIP_SERVERS);
                    Set<String> skipListSet = null;
                    if (skipList != null) {
                        skipListSet = new HashSet<>(skipList.length);
                        skipListSet.addAll(Arrays.asList(skipList));
                        log.info("test.distrib.skip.servers was found and contains:" + skipListSet);
                    }
                    for (ZkCoreNodeProps props : replicaProps) {
                        if (skipList != null) {
                            boolean skip = skipListSet.contains(props.getCoreUrl());
                            log.info("check url:" + props.getCoreUrl() + " against:" + skipListSet + " result:" + skip);
                            if (!skip) {
                                nodes.add(new StdNode(props, collection, shardId));
                            }
                        } else {
                            nodes.add(new StdNode(props, collection, shardId));
                        }
                    }
                }
            } else {
                // I need to forward onto the leader...
                nodes = new ArrayList<>(1);
                nodes.add(new RetryNode(new ZkCoreNodeProps(leaderReplica), zkController.getZkStateReader(), collection, shardId));
                forwardToLeader = true;
            }
        } catch (InterruptedException e) {
            Thread.currentThread().interrupt();
            throw new ZooKeeperException(SolrException.ErrorCode.SERVER_ERROR, "", e);
        }
    }
    return nodes;
}

Example 4 with RetryNode

use of org.apache.solr.update.SolrCmdDistributor.RetryNode in project lucene-solr by apache.

the class SolrCmdDistributorTest method testRetryNodeWontRetrySocketError.

private void testRetryNodeWontRetrySocketError() throws Exception {
    final HttpSolrClient solrclient = (HttpSolrClient) clients.get(0);
    long numFoundBefore = solrclient.query(new SolrQuery("*:*")).getResults().getNumFound();
    final MockStreamingSolrClients streamingClients = new MockStreamingSolrClients(updateShardHandler);
    SolrCmdDistributor cmdDistrib = new SolrCmdDistributor(streamingClients, 5, 0);
    streamingClients.setExp(Exp.SOCKET_EXCEPTION);
    ArrayList<Node> nodes = new ArrayList<>();
    ZkNodeProps nodeProps = new ZkNodeProps(ZkStateReader.BASE_URL_PROP, solrclient.getBaseURL(), ZkStateReader.CORE_NAME_PROP, "");
    final AtomicInteger retries = new AtomicInteger();
    nodeProps = new ZkNodeProps(ZkStateReader.BASE_URL_PROP, solrclient.getBaseURL(), ZkStateReader.CORE_NAME_PROP, "");
    RetryNode retryNode = new RetryNode(new ZkCoreNodeProps(nodeProps), null, "collection1", "shard1") {

        @Override
        public boolean checkRetry() {
            retries.incrementAndGet();
            return true;
        }
    };
    nodes.add(retryNode);
    AddUpdateCommand cmd = new AddUpdateCommand(null);
    cmd.solrDoc = sdoc("id", id.incrementAndGet());
    ModifiableSolrParams params = new ModifiableSolrParams();
    CommitUpdateCommand ccmd = new CommitUpdateCommand(null, false);
    cmdDistrib.distribAdd(cmd, nodes, params);
    streamingClients.setExp(null);
    cmdDistrib.distribCommit(ccmd, nodes, params);
    cmdDistrib.finish();
    // it will checkRetry, but not actually do it...
    assertEquals(1, retries.get());
    long numFoundAfter = solrclient.query(new SolrQuery("*:*")).getResults().getNumFound();
    // we will get java.net.SocketException: Network is unreachable, which we don't retry on
    assertEquals(numFoundBefore, numFoundAfter);
    assertEquals(1, cmdDistrib.getErrors().size());
}

Example 5 with RetryNode

use of org.apache.solr.update.SolrCmdDistributor.RetryNode in project lucene-solr by apache.

the class SolrCmdDistributorTest method testRetryNodeAgainstBadAddress.

private void testRetryNodeAgainstBadAddress() throws SolrServerException, IOException {
    // Test RetryNode
    SolrCmdDistributor cmdDistrib = new SolrCmdDistributor(updateShardHandler);
    final HttpSolrClient solrclient = (HttpSolrClient) clients.get(0);
    long numFoundBefore = solrclient.query(new SolrQuery("*:*")).getResults().getNumFound();
    ArrayList<Node> nodes = new ArrayList<>();
    ZkNodeProps nodeProps = new ZkNodeProps(ZkStateReader.BASE_URL_PROP, "[ff01::114]:33332" + context, ZkStateReader.CORE_NAME_PROP, "");
    RetryNode retryNode = new RetryNode(new ZkCoreNodeProps(nodeProps), null, "collection1", "shard1") {

        @Override
        public boolean checkRetry() {
            ZkNodeProps leaderProps = new ZkNodeProps(ZkStateReader.BASE_URL_PROP, solrclient.getBaseURL(), ZkStateReader.CORE_NAME_PROP, "");
            this.nodeProps = new ZkCoreNodeProps(leaderProps);
            return true;
        }
    };
    nodes.add(retryNode);
    AddUpdateCommand cmd = new AddUpdateCommand(null);
    cmd.solrDoc = sdoc("id", id.incrementAndGet());
    ModifiableSolrParams params = new ModifiableSolrParams();
    cmdDistrib.distribAdd(cmd, nodes, params);
    CommitUpdateCommand ccmd = new CommitUpdateCommand(null, false);
    params = new ModifiableSolrParams();
    params.set(DistributedUpdateProcessor.COMMIT_END_POINT, true);
    cmdDistrib.distribCommit(ccmd, nodes, params);
    cmdDistrib.finish();
    long numFoundAfter = solrclient.query(new SolrQuery("*:*")).getResults().getNumFound();
    // different OS's will throw different exceptions for the bad address above
    if (numFoundBefore != numFoundAfter) {
        assertEquals(0, cmdDistrib.getErrors().size());
        assertEquals(numFoundBefore + 1, numFoundAfter);
    } else {
        // we will get java.net.SocketException: Network is unreachable and not retry
        assertEquals(numFoundBefore, numFoundAfter);
        assertEquals(1, cmdDistrib.getErrors().size());
    }
}