Search in sources :

Example 21 with Update

use of io.cdap.cdap.data2.dataset2.lib.table.Update in project cdap by caskdata.

the class HBaseTableFactory method upgradeCoProcessor.

private void upgradeCoProcessor(TableId tableId, Class<? extends Coprocessor> coprocessor) throws IOException {
    try (HBaseDDLExecutor ddlExecutor = ddlExecutorFactory.get()) {
        HTableDescriptor tableDescriptor;
        try (HBaseAdmin admin = new HBaseAdmin(hConf)) {
            // If table doesn't exist, then skip upgrading coprocessor
            if (!tableUtil.tableExists(admin, tableId)) {
                LOG.debug("TMS Table {} was not found. Skip upgrading coprocessor.", tableId);
                return;
            }
            tableDescriptor = tableUtil.getHTableDescriptor(admin, tableId);
        }
        // Get cdap version from the table
        ProjectInfo.Version version = HBaseTableUtil.getVersion(tableDescriptor);
        String hbaseVersion = HBaseTableUtil.getHBaseVersion(tableDescriptor);
        if (hbaseVersion != null && hbaseVersion.equals(HBaseVersion.getVersionString()) && version.compareTo(ProjectInfo.getVersion()) >= 0) {
            // If cdap has version has not changed or is greater, no need to update. Just enable it, in case
            // it has been disabled by the upgrade tool, and return
            LOG.info("Table '{}' has not changed and its version '{}' is same or greater than current CDAP version '{}'." + " The underlying HBase version {} has also not changed.", tableId, version, ProjectInfo.getVersion(), hbaseVersion);
            enableTable(ddlExecutor, tableId);
            return;
        }
        // create a new descriptor for the table update
        HTableDescriptorBuilder newDescriptor = tableUtil.buildHTableDescriptor(tableDescriptor);
        // Remove old coprocessor
        Map<String, HBaseTableUtil.CoprocessorInfo> coprocessorInfo = HBaseTableUtil.getCoprocessorInfo(tableDescriptor);
        for (Map.Entry<String, HBaseTableUtil.CoprocessorInfo> coprocessorEntry : coprocessorInfo.entrySet()) {
            newDescriptor.removeCoprocessor(coprocessorEntry.getValue().getClassName());
        }
        // Add new coprocessor
        CoprocessorDescriptor coprocessorDescriptor = coprocessorManager.getCoprocessorDescriptor(coprocessor, Coprocessor.PRIORITY_USER);
        Path path = coprocessorDescriptor.getPath() == null ? null : new Path(coprocessorDescriptor.getPath());
        newDescriptor.addCoprocessor(coprocessorDescriptor.getClassName(), path, coprocessorDescriptor.getPriority(), coprocessorDescriptor.getProperties());
        // Update CDAP version, table prefix
        HBaseTableUtil.setVersion(newDescriptor);
        HBaseTableUtil.setHBaseVersion(newDescriptor);
        HBaseTableUtil.setTablePrefix(newDescriptor, cConf);
        // Disable auto-splitting
        newDescriptor.setValue(HTableDescriptor.SPLIT_POLICY, cConf.get(Constants.MessagingSystem.TABLE_HBASE_SPLIT_POLICY));
        // Disable Table
        disableTable(ddlExecutor, tableId);
        tableUtil.modifyTable(ddlExecutor, newDescriptor.build());
        LOG.debug("Enabling table '{}'...", tableId);
        enableTable(ddlExecutor, tableId);
    }
    LOG.info("Table '{}' update completed.", tableId);
}
Also used : HBaseDDLExecutor(io.cdap.cdap.spi.hbase.HBaseDDLExecutor) Path(org.apache.hadoop.fs.Path) HTableDescriptorBuilder(io.cdap.cdap.data2.util.hbase.HTableDescriptorBuilder) HTableDescriptor(org.apache.hadoop.hbase.HTableDescriptor) HBaseAdmin(org.apache.hadoop.hbase.client.HBaseAdmin) ProjectInfo(io.cdap.cdap.common.utils.ProjectInfo) CoprocessorDescriptor(io.cdap.cdap.spi.hbase.CoprocessorDescriptor) Map(java.util.Map) ConcurrentHashMap(java.util.concurrent.ConcurrentHashMap)

Example 22 with Update

use of io.cdap.cdap.data2.dataset2.lib.table.Update in project cdap by caskdata.

the class AppMetadataStore method addWorkflowNodeState.

private void addWorkflowNodeState(ProgramId programId, String pid, Map<String, String> systemArgs, ProgramRunStatus status, @Nullable BasicThrowable failureCause) {
    String workflowNodeId = systemArgs.get(ProgramOptionConstants.WORKFLOW_NODE_ID);
    String workflowName = systemArgs.get(ProgramOptionConstants.WORKFLOW_NAME);
    String workflowRun = systemArgs.get(ProgramOptionConstants.WORKFLOW_RUN_ID);
    ApplicationId appId = Ids.namespace(programId.getNamespace()).app(programId.getApplication());
    ProgramRunId workflowRunId = appId.workflow(workflowName).run(workflowRun);
    // Node states will be stored with following key:
    // workflowNodeState.namespace.app.WORKFLOW.workflowName.workflowRun.workflowNodeId
    MDSKey key = getProgramKeyBuilder(TYPE_WORKFLOW_NODE_STATE, workflowRunId.getParent()).add(workflowRun).add(workflowNodeId).build();
    WorkflowNodeStateDetail nodeStateDetail = new WorkflowNodeStateDetail(workflowNodeId, ProgramRunStatus.toNodeStatus(status), pid, failureCause);
    write(key, nodeStateDetail);
    // Get the run record of the Workflow which started this program
    key = getProgramKeyBuilder(TYPE_RUN_RECORD_STARTED, workflowRunId.getParent()).add(workflowRunId.getRun()).build();
    RunRecordMeta record = get(key, RunRecordMeta.class);
    if (record != null) {
        // Update the parent Workflow run record by adding node id and program run id in the properties
        Map<String, String> properties = record.getProperties();
        properties.put(workflowNodeId, pid);
        write(key, new RunRecordMeta(record, properties));
    }
}
Also used : MDSKey(co.cask.cdap.data2.dataset2.lib.table.MDSKey) ProgramRunId(co.cask.cdap.proto.id.ProgramRunId) ApplicationId(co.cask.cdap.proto.id.ApplicationId) WorkflowNodeStateDetail(co.cask.cdap.proto.WorkflowNodeStateDetail)

Example 23 with Update

use of io.cdap.cdap.data2.dataset2.lib.table.Update in project cdap by caskdata.

the class DatasetTypeManager method addModule.

/**
 * Add datasets module in a namespace
 *
 * @param datasetModuleId the {@link DatasetModuleId} to add
 * @param className module class
 * @param jarLocation location of the module jar
 * @param force if true, an update will be allowed even if there are conflicts with other modules, or if
 *                     removal of a type would break other modules' dependencies.
 */
public void addModule(final DatasetModuleId datasetModuleId, final String className, final Location jarLocation, final boolean force) throws DatasetModuleConflictException {
    LOG.debug("adding module: {}, className: {}, jarLocation: {}", datasetModuleId, className, jarLocation == null ? "[local]" : jarLocation);
    try {
        final DatasetTypeMDS datasetTypeMDS = datasetCache.getDataset(DatasetMetaTableUtil.META_TABLE_NAME);
        final DatasetInstanceMDS datasetInstanceMDS = datasetCache.getDataset(DatasetMetaTableUtil.INSTANCE_TABLE_NAME);
        txExecutorFactory.createExecutor(datasetCache).execute(new TransactionExecutor.Subroutine() {

            @Override
            public void apply() throws Exception {
                // 1. get existing module with all its types
                DatasetModuleMeta existing = datasetTypeMDS.getModule(datasetModuleId);
                DependencyTrackingRegistry reg;
                // 2. unpack jar and create class loader
                File unpackedLocation = Files.createTempDirectory(Files.createDirectories(systemTempPath), datasetModuleId.getEntityName()).toFile();
                DirectoryClassLoader cl = null;
                try {
                    // NOTE: if jarLocation is null, we assume that this is a system module, ie. always present in classpath
                    if (jarLocation != null) {
                        BundleJarUtil.unJar(jarLocation, unpackedLocation);
                        cl = new DirectoryClassLoader(unpackedLocation, cConf.get(Constants.AppFabric.PROGRAM_EXTRA_CLASSPATH), FilterClassLoader.create(getClass().getClassLoader()), "lib");
                    }
                    reg = new DependencyTrackingRegistry(datasetModuleId, datasetTypeMDS, cl, force);
                    // 3. register the new module while tracking dependencies.
                    // this will fail if a type exists in a different module
                    DatasetDefinitionRegistries.register(className, cl, reg);
                } catch (TypeConflictException e) {
                    // type conflict from the registry, we want to throw that as is
                    throw e;
                } catch (Exception e) {
                    LOG.error("Could not instantiate instance of dataset module class {} for module {} using jarLocation {}", className, datasetModuleId, jarLocation);
                    throw Throwables.propagate(e);
                } finally {
                    if (cl != null) {
                        // Close the ProgramClassLoader
                        Closeables.closeQuietly(cl);
                    }
                    try {
                        DirUtils.deleteDirectoryContents(unpackedLocation);
                    } catch (IOException e) {
                        LOG.warn("Failed to delete directory {}", unpackedLocation, e);
                    }
                }
                // 4. determine whether any type were removed from the module, and whether any other modules depend on them
                if (existing != null) {
                    Set<String> removedTypes = new HashSet<>(existing.getTypes());
                    removedTypes.removeAll(reg.getTypes());
                    // TODO (CDAP-6294): track dependencies at the type level
                    if (!force && !removedTypes.isEmpty() && !existing.getUsedByModules().isEmpty()) {
                        throw new DatasetModuleConflictException(String.format("Cannot update module '%s' to remove types %s: Modules %s may depend on it. Delete them first", datasetModuleId, removedTypes, existing.getUsedByModules()));
                    }
                    Collection<DatasetSpecification> instances = datasetInstanceMDS.getByTypes(datasetModuleId.getParent(), removedTypes);
                    if (!instances.isEmpty()) {
                        throw new DatasetModuleConflictException(String.format("Attempt to remove dataset types %s from module '%s' that have existing instances: %s. " + "Delete them first.", removedTypes, datasetModuleId, Iterables.toString(Iterables.transform(instances, new Function<DatasetSpecification, String>() {

                            @Nullable
                            @Override
                            public String apply(@Nullable DatasetSpecification input) {
                                return input.getName() + ":" + input.getType();
                            }
                        }))));
                    }
                }
                // NOTE: we use set to avoid duplicated dependencies
                // NOTE: we use LinkedHashSet to preserve order in which dependencies must be loaded
                Set<String> moduleDependencies = new LinkedHashSet<String>();
                for (DatasetTypeId usedType : reg.getUsedTypes()) {
                    DatasetModuleMeta usedModule = datasetTypeMDS.getModuleByType(usedType);
                    Preconditions.checkState(usedModule != null, String.format("Found a null used module for type %s for while adding module %s", usedType, datasetModuleId));
                    // adding all used types and the module itself, in this very order to keep the order of loading modules
                    // for instantiating a type
                    moduleDependencies.addAll(usedModule.getUsesModules());
                    boolean added = moduleDependencies.add(usedModule.getName());
                    if (added) {
                        // also adding this module as a dependent for all modules it uses
                        usedModule.addUsedByModule(datasetModuleId.getEntityName());
                        datasetTypeMDS.writeModule(usedType.getParent(), usedModule);
                    }
                }
                URI jarURI = jarLocation == null ? null : jarLocation.toURI();
                DatasetModuleMeta moduleMeta = existing == null ? new DatasetModuleMeta(datasetModuleId.getEntityName(), className, jarURI, reg.getTypes(), Lists.newArrayList(moduleDependencies)) : new DatasetModuleMeta(datasetModuleId.getEntityName(), className, jarURI, reg.getTypes(), Lists.newArrayList(moduleDependencies), Lists.newArrayList(existing.getUsedByModules()));
                datasetTypeMDS.writeModule(datasetModuleId.getParent(), moduleMeta);
            }
        });
    } catch (TransactionFailureException e) {
        Throwable cause = e.getCause();
        if (cause != null) {
            if (cause instanceof DatasetModuleConflictException) {
                throw (DatasetModuleConflictException) cause;
            } else if (cause instanceof TypeConflictException) {
                throw new DatasetModuleConflictException(cause.getMessage(), cause);
            }
        }
        throw Throwables.propagate(e);
    } catch (Exception e) {
        LOG.error("Operation failed", e);
        throw Throwables.propagate(e);
    }
}
Also used : ImmutableSet(com.google.common.collect.ImmutableSet) Set(java.util.Set) HashSet(java.util.HashSet) LinkedHashSet(java.util.LinkedHashSet) DatasetTypeId(co.cask.cdap.proto.id.DatasetTypeId) URI(java.net.URI) DatasetModuleMeta(co.cask.cdap.proto.DatasetModuleMeta) DirectoryClassLoader(co.cask.cdap.common.lang.DirectoryClassLoader) TypeConflictException(co.cask.cdap.data2.dataset2.TypeConflictException) DatasetInstanceMDS(co.cask.cdap.data2.datafabric.dataset.service.mds.DatasetInstanceMDS) DatasetSpecification(co.cask.cdap.api.dataset.DatasetSpecification) TransactionExecutor(org.apache.tephra.TransactionExecutor) IOException(java.io.IOException) TransactionFailureException(org.apache.tephra.TransactionFailureException) TypeConflictException(co.cask.cdap.data2.dataset2.TypeConflictException) IOException(java.io.IOException) DatasetTypeMDS(co.cask.cdap.data2.datafabric.dataset.service.mds.DatasetTypeMDS) TransactionFailureException(org.apache.tephra.TransactionFailureException) Collection(java.util.Collection) File(java.io.File) Nullable(javax.annotation.Nullable)

Example 24 with Update

use of io.cdap.cdap.data2.dataset2.lib.table.Update in project cdap by caskdata.

the class InMemoryTableServiceTest method testInternalsNotLeaking.

@Test
public void testInternalsNotLeaking() {
    // Test that there's no way to break the state of InMemoryTableService by changing parameters of update
    // methods (after method invocation) or by changing returned values "in-place"
    InMemoryTableService.create("table");
    // verify writing thru merge is guarded
    NavigableMap<byte[], NavigableMap<byte[], Update>> updates = Maps.newTreeMap(Bytes.BYTES_COMPARATOR);
    NavigableMap<byte[], Update> rowUpdate = Maps.newTreeMap(Bytes.BYTES_COMPARATOR);
    byte[] rowParam = new byte[] { 1 };
    byte[] columnParam = new byte[] { 2 };
    byte[] valParam = new byte[] { 3 };
    rowUpdate.put(columnParam, new PutValue(valParam));
    updates.put(rowParam, rowUpdate);
    InMemoryTableService.merge("table", updates, 1L);
    verify123();
    updates.remove(rowParam);
    rowUpdate.remove(columnParam);
    rowParam[0]++;
    columnParam[0]++;
    valParam[0]++;
    verify123();
    // verify changing returned data from get doesn't affect the stored data
    NavigableMap<byte[], NavigableMap<Long, byte[]>> rowFromGet = InMemoryTableService.get("table", new byte[] { 1 }, new Transaction(1L, 2L, new long[0], new long[0], 1L));
    Assert.assertEquals(1, rowFromGet.size());
    byte[] columnFromGet = rowFromGet.firstEntry().getKey();
    Assert.assertArrayEquals(new byte[] { 2 }, columnFromGet);
    byte[] valFromGet = rowFromGet.firstEntry().getValue().get(1L);
    Assert.assertArrayEquals(new byte[] { 3 }, valFromGet);
    rowFromGet.firstEntry().getValue().remove(1L);
    rowFromGet.remove(columnFromGet);
    columnFromGet[0]++;
    valFromGet[0]++;
    verify123();
    // verify changing returned data doesn't affect the stored data
    NavigableMap<byte[], NavigableMap<byte[], NavigableMap<Long, byte[]>>> fromGetRange = InMemoryTableService.getRowRange("table", null, null, new Transaction(1L, 2L, new long[0], new long[0], 1L));
    Assert.assertEquals(1, fromGetRange.size());
    byte[] keyFromGetRange = fromGetRange.firstEntry().getKey();
    Assert.assertArrayEquals(new byte[] { 1 }, keyFromGetRange);
    NavigableMap<byte[], NavigableMap<Long, byte[]>> rowFromGetRange = fromGetRange.get(new byte[] { 1 });
    Assert.assertEquals(1, rowFromGetRange.size());
    byte[] columnFromGetRange = rowFromGetRange.firstEntry().getKey();
    Assert.assertArrayEquals(new byte[] { 2 }, columnFromGetRange);
    byte[] valFromGetRange = rowFromGetRange.firstEntry().getValue().get(1L);
    Assert.assertArrayEquals(new byte[] { 3 }, valFromGetRange);
    rowFromGetRange.firstEntry().getValue().remove(1L);
    rowFromGetRange.remove(columnFromGetRange);
    fromGetRange.remove(keyFromGetRange);
    keyFromGetRange[0]++;
    columnFromGetRange[0]++;
    valFromGet[0]++;
    verify123();
}
Also used : PutValue(co.cask.cdap.data2.dataset2.lib.table.PutValue) NavigableMap(java.util.NavigableMap) Transaction(org.apache.tephra.Transaction) Update(co.cask.cdap.data2.dataset2.lib.table.Update) Test(org.junit.Test)

Example 25 with Update

use of io.cdap.cdap.data2.dataset2.lib.table.Update in project cdap by caskdata.

the class HBaseTable method undo.

@Override
protected void undo(NavigableMap<byte[], NavigableMap<byte[], Update>> persisted) throws Exception {
    if (persisted.isEmpty()) {
        return;
    }
    // NOTE: we use Delete with the write pointer as the specific version to delete.
    List<Delete> deletes = Lists.newArrayList();
    for (Map.Entry<byte[], NavigableMap<byte[], Update>> row : persisted.entrySet()) {
        DeleteBuilder delete = tableUtil.buildDelete(row.getKey());
        delete.setAttribute(TX_MAX_LIFETIME_MILLIS_KEY, txMaxLifetimeMillis);
        for (Map.Entry<byte[], Update> column : row.getValue().entrySet()) {
            // we want support tx and non-tx modes
            if (tx != null) {
                delete.setAttribute(TxConstants.TX_ROLLBACK_ATTRIBUTE_KEY, new byte[0]);
                // TODO: hijacking timestamp... bad
                delete.deleteColumn(columnFamily, column.getKey(), tx.getWritePointer());
            } else {
                delete.deleteColumns(columnFamily, column.getKey());
            }
        }
        deletes.add(delete.build());
    }
    if (!deletes.isEmpty()) {
        hbaseDelete(deletes);
    }
}
Also used : Delete(org.apache.hadoop.hbase.client.Delete) NavigableMap(java.util.NavigableMap) Update(co.cask.cdap.data2.dataset2.lib.table.Update) Map(java.util.Map) ImmutableMap(com.google.common.collect.ImmutableMap) NavigableMap(java.util.NavigableMap) DeleteBuilder(co.cask.cdap.data2.util.hbase.DeleteBuilder)

Aggregations

NavigableMap (java.util.NavigableMap)15 Map (java.util.Map)13 Test (org.junit.Test)8 Update (co.cask.cdap.data2.dataset2.lib.table.Update)7 Update (io.cdap.cdap.data2.dataset2.lib.table.Update)7 DatasetSpecification (io.cdap.cdap.api.dataset.DatasetSpecification)6 IOException (java.io.IOException)6 ArrayList (java.util.ArrayList)6 ConcurrentNavigableMap (java.util.concurrent.ConcurrentNavigableMap)6 PutValue (co.cask.cdap.data2.dataset2.lib.table.PutValue)4 PutValue (io.cdap.cdap.data2.dataset2.lib.table.PutValue)4 Collection (java.util.Collection)4 HashSet (java.util.HashSet)4 Set (java.util.Set)4 DatasetManagementException (co.cask.cdap.api.dataset.DatasetManagementException)3 IncrementValue (co.cask.cdap.data2.dataset2.lib.table.IncrementValue)3 SortedMap (java.util.SortedMap)3 ConcurrentSkipListMap (java.util.concurrent.ConcurrentSkipListMap)3 Table (co.cask.cdap.api.dataset.table.Table)2 InMemoryTableModule (co.cask.cdap.data2.dataset2.module.lib.inmemory.InMemoryTableModule)2