Examples with DataMediaPair com.alibaba.otter.shared.common.model.config.data.DataMediaPair used on opensource projects

Example 36 with DataMediaPair

use of com.alibaba.otter.shared.common.model.config.data.DataMediaPair in project otter by alibaba.

the class DataMediaPairInfo method execute.

public void execute(@Param("dataMediaPairId") Long dataMediaPairId, Context context) throws Exception {
    DataMediaPair dataMediaPair = dataMediaPairService.findById(dataMediaPairId);
    Channel channel = channelService.findByPipelineId(dataMediaPair.getPipelineId());
    List<ColumnPair> columnPairs = dataMediaPair.getColumnPairs();
    List<ColumnGroup> columnGroups = dataMediaPair.getColumnGroups();
    // 暂时策略，只拿出list的第一个Group
    ColumnGroup columnGroup = new ColumnGroup();
    if (!CollectionUtils.isEmpty(columnGroups)) {
        columnGroup = columnGroups.get(0);
    }
    context.put("dataMediaPair", dataMediaPair);
    context.put("columnGroup", columnGroup);
    context.put("columnPairs", columnPairs);
    context.put("channelId", channel.getId());
}

Example 37 with DataMediaPair

use of com.alibaba.otter.shared.common.model.config.data.DataMediaPair in project otter by alibaba.

the class EditDataMediaPair method execute.

public void execute(@Param("dataMediaPairId") Long dataMediaPairId, @Param("pipelineId") Long pipelineId, Context context, Navigator nav) throws Exception {
    Channel channel = channelService.findByPipelineId(pipelineId);
    if (channel.getStatus().isStart()) {
        nav.redirectTo(WebConstant.ERROR_FORBIDDEN_Link);
        return;
    }
    DataMediaPair dataMediaPair = dataMediaPairService.findById(dataMediaPairId);
    context.put("channelId", channel.getId());
    context.put("dataMediaPair", dataMediaPair);
}

Example 38 with DataMediaPair

use of com.alibaba.otter.shared.common.model.config.data.DataMediaPair in project otter by alibaba.

the class ProcessorExtractor method extract.

public void extract(DbBatch param) throws ExtractException {
    ExecutorTemplate executorTemplate = null;
    try {
        RowBatch rowBatch = param.getRowBatch();
        final Pipeline pipeline = getPipeline(rowBatch.getIdentity().getPipelineId());
        List<EventData> eventDatas = rowBatch.getDatas();
        // 使用set，提升remove时的查找速度
        final Set<EventData> removeDatas = Collections.synchronizedSet(new HashSet<EventData>());
        executorTemplate = executorTemplateGetter.get();
        executorTemplate.start();
        // 重新设置下poolSize
        executorTemplate.adjustPoolSize(pipeline.getParameters().getExtractPoolSize());
        for (final EventData eventData : eventDatas) {
            List<DataMediaPair> dataMediaPairs = ConfigHelper.findDataMediaPairByMediaId(pipeline, eventData.getTableId());
            if (dataMediaPairs == null) {
                throw new ExtractException("ERROR ## the dataMediaId = " + eventData.getTableId() + " dataMediaPair is null,please check");
            }
            for (DataMediaPair dataMediaPair : dataMediaPairs) {
                if (!dataMediaPair.isExistFilter()) {
                    continue;
                }
                final EventProcessor eventProcessor = extensionFactory.getExtension(EventProcessor.class, dataMediaPair.getFilterData());
                if (eventProcessor instanceof DataSourceFetcherAware) {
                    ((DataSourceFetcherAware) eventProcessor).setDataSourceFetcher(new DataSourceFetcher() {

                        @Override
                        public DataSource fetch(Long tableId) {
                            DataMedia dataMedia = ConfigHelper.findDataMedia(pipeline, tableId);
                            return dataSourceService.getDataSource(pipeline.getId(), dataMedia.getSource());
                        }
                    });
                    executorTemplate.submit(new Runnable() {

                        @Override
                        public void run() {
                            MDC.put(OtterConstants.splitPipelineLogFileKey, String.valueOf(pipeline.getId()));
                            boolean process = eventProcessor.process(eventData);
                            if (!process) {
                                // 添加到删除记录中
                                removeDatas.add(eventData);
                            }
                        }
                    });
                } else {
                    boolean process = eventProcessor.process(eventData);
                    if (!process) {
                        // 添加到删除记录中
                        removeDatas.add(eventData);
                        break;
                    }
                }
            }
        }
        // 等待所有都处理完成
        executorTemplate.waitForResult();
        if (!CollectionUtils.isEmpty(removeDatas)) {
            eventDatas.removeAll(removeDatas);
        }
    } finally {
        if (executorTemplate != null) {
            executorTemplateGetter.release(executorTemplate);
        }
    }
}

Example 39 with DataMediaPair

use of com.alibaba.otter.shared.common.model.config.data.DataMediaPair in project otter by alibaba.

the class ViewExtractor method extract.

@Override
public void extract(DbBatch dbBatch) throws ExtractException {
    Assert.notNull(dbBatch);
    Assert.notNull(dbBatch.getRowBatch());
    Pipeline pipeline = getPipeline(dbBatch.getRowBatch().getIdentity().getPipelineId());
    List<DataMediaPair> dataMediaPairs = pipeline.getPairs();
    /**
         * Key = TableId<br>
         * Value = a List of this tableId's column need to sync<br>
         */
    Map<Long, List<ColumnPair>> viewColumnPairs = new HashMap<Long, List<ColumnPair>>();
    Map<Long, ColumnPairMode> viewColumnPairModes = new HashMap<Long, ColumnPairMode>();
    for (DataMediaPair dataMediaPair : dataMediaPairs) {
        List<ColumnPair> columnPairs = dataMediaPair.getColumnPairs();
        // 设置ColumnPairMode
        viewColumnPairModes.put(dataMediaPair.getSource().getId(), dataMediaPair.getColumnPairMode());
        // 如果没有columnPairs，则默认全字段同步，不做处理
        if (!CollectionUtils.isEmpty(columnPairs)) {
            viewColumnPairs.put(dataMediaPair.getSource().getId(), columnPairs);
        }
    }
    List<EventData> eventDatas = dbBatch.getRowBatch().getDatas();
    // 使用set，提升remove时的查找速度
    Set<EventData> removeDatas = new HashSet<EventData>();
    for (EventData eventData : eventDatas) {
        if (eventData.getEventType().isDdl()) {
            continue;
        }
        List<ColumnPair> columns = viewColumnPairs.get(eventData.getTableId());
        if (!CollectionUtils.isEmpty(columns)) {
            // 组装需要同步的Column
            ColumnPairMode mode = viewColumnPairModes.get(eventData.getTableId());
            eventData.setColumns(columnFilter(eventData.getColumns(), columns, mode));
            eventData.setKeys(columnFilter(eventData.getKeys(), columns, mode));
            if (!CollectionUtils.isEmpty(eventData.getOldKeys())) {
                eventData.setOldKeys(columnFilter(eventData.getOldKeys(), columns, mode));
            }
            if (CollectionUtils.isEmpty(eventData.getKeys())) {
                // 无主键，报错
                throw new ExtractException(String.format("eventData after viewExtractor has no pks , pls check! identity:%s, new eventData:%s", dbBatch.getRowBatch().getIdentity().toString(), eventData.toString()));
            }
            // update: 过滤后如果无字段(变更需要同步)和主键变更，则可以忽略之，避免sql语法错误
            if (eventData.getEventType().isUpdate() && (CollectionUtils.isEmpty(eventData.getColumns()) || CollectionUtils.isEmpty(eventData.getUpdatedColumns())) && CollectionUtils.isEmpty(eventData.getOldKeys())) {
                // 过滤之后无字段需要同步，并且不存在主键变更同步，则忽略该记录
                removeDatas.add(eventData);
            }
        }
    }
    if (!CollectionUtils.isEmpty(removeDatas)) {
        eventDatas.removeAll(removeDatas);
    }
}

Example 40 with DataMediaPair

use of com.alibaba.otter.shared.common.model.config.data.DataMediaPair in project otter by alibaba.

the class MessageParser method internParse.

/**
     * 解析出从canal中获取的Event事件<br>
     * Oracle:有变更的列值. <br>
     * <i>insert:从afterColumns中获取所有的变更数据<br>
     * <i>delete:从beforeColumns中获取所有的变更数据<br>
     * <i>update:在before中存放所有的主键和变化前的非主键值，在after中存放变化后的主键和非主键值,如果是复合主键，只会存放变化的主键<br>
     * Mysql:可以得到所有变更前和变更后的数据.<br>
     * <i>insert:从afterColumns中获取所有的变更数据<br>
     * <i>delete:从beforeColumns中获取所有的变更数据<br>
     * <i>update:在beforeColumns中存放变更前的所有数据,在afterColumns中存放变更后的所有数据<br>
     */
private EventData internParse(Pipeline pipeline, Entry entry, RowChange rowChange, RowData rowData) {
    EventData eventData = new EventData();
    eventData.setTableName(entry.getHeader().getTableName());
    eventData.setSchemaName(entry.getHeader().getSchemaName());
    eventData.setEventType(EventType.valueOf(rowChange.getEventType().name()));
    eventData.setExecuteTime(entry.getHeader().getExecuteTime());
    EventType eventType = eventData.getEventType();
    TableInfoHolder tableHolder = null;
    if (!StringUtils.equalsIgnoreCase(pipeline.getParameters().getSystemSchema(), eventData.getSchemaName())) {
        boolean useTableTransform = pipeline.getParameters().getUseTableTransform();
        Table table = null;
        DataMediaPair dataMediaPair = ConfigHelper.findDataMediaPairBySourceName(pipeline, eventData.getSchemaName(), eventData.getTableName());
        DataMedia dataMedia = dataMediaPair.getSource();
        eventData.setTableId(dataMedia.getId());
        // 获取目标表
        DataMedia targetDataMedia = dataMediaPair.getTarget();
        if (useTableTransform || dataMedia.getSource().getType().isOracle()) {
            // oracle需要反查一次meta
            // 如果设置了需要进行table meta转化，则反查一下table信息
            // 比如oracle erosa解析时可能使用了非物理主键，需要直接使用，信任erosa的信息
            DbDialect dbDialect = dbDialectFactory.getDbDialect(pipeline.getId(), (DbMediaSource) dataMedia.getSource());
            // 查询一下meta信息
            table = dbDialect.findTable(eventData.getSchemaName(), eventData.getTableName());
            if (table == null) {
                logger.warn("find table[{}.{}] is null , may be drop table.", eventData.getSchemaName(), eventData.getTableName());
            }
            // 获取一下目标库的拆分字段,设置源表为主键
            // 首先要求源和目标的库名表名是一致的
            DbDialect targetDbDialect = dbDialectFactory.getDbDialect(pipeline.getId(), (DbMediaSource) targetDataMedia.getSource());
            if (targetDbDialect.isDRDS()) {
                String schemaName = buildName(eventData.getSchemaName(), dataMedia.getNamespaceMode(), targetDataMedia.getNamespaceMode());
                String tableName = buildName(eventData.getSchemaName(), dataMedia.getNameMode(), targetDataMedia.getNameMode());
                String shardColumns = targetDbDialect.getShardColumns(schemaName, tableName);
                if (StringUtils.isNotEmpty(shardColumns)) {
                    String[] columns = StringUtils.split(shardColumns, ',');
                    for (String key : columns) {
                        org.apache.ddlutils.model.Column col = table.findColumn(key, false);
                        if (col != null) {
                            col.setPrimaryKey(true);
                        } else {
                            logger.warn(String.format("shardColumn %s in table[%s.%s] is not found", key, eventData.getSchemaName(), eventData.getTableName()));
                        }
                    }
                }
            }
            tableHolder = new TableInfoHolder(dbDialect, table, useTableTransform);
        }
    }
    List<Column> beforeColumns = rowData.getBeforeColumnsList();
    List<Column> afterColumns = rowData.getAfterColumnsList();
    String tableName = eventData.getSchemaName() + "." + eventData.getTableName();
    // 判断一下是否需要all columns
    // 如果是rowMode模式，所有字段都需要标记为updated
    boolean isRowMode = pipeline.getParameters().getSyncMode().isRow();
    boolean needAllColumns = isRowMode || checkNeedAllColumns(pipeline);
    // 变更后的主键
    Map<String, EventColumn> keyColumns = new LinkedHashMap<String, EventColumn>();
    // 变更前的主键
    Map<String, EventColumn> oldKeyColumns = new LinkedHashMap<String, EventColumn>();
    // 有变化的非主键
    Map<String, EventColumn> notKeyColumns = new LinkedHashMap<String, EventColumn>();
    if (eventType.isInsert()) {
        for (Column column : afterColumns) {
            if (isKey(tableHolder, tableName, column)) {
                keyColumns.put(column.getName(), copyEventColumn(column, true, tableHolder));
            } else {
                // mysql 有效
                notKeyColumns.put(column.getName(), copyEventColumn(column, true, tableHolder));
            }
        }
    } else if (eventType.isDelete()) {
        for (Column column : beforeColumns) {
            if (isKey(tableHolder, tableName, column)) {
                keyColumns.put(column.getName(), copyEventColumn(column, true, tableHolder));
            } else {
                // mysql 有效
                notKeyColumns.put(column.getName(), copyEventColumn(column, true, tableHolder));
            }
        }
    } else if (eventType.isUpdate()) {
        // 获取变更前的主键.
        for (Column column : beforeColumns) {
            if (isKey(tableHolder, tableName, column)) {
                oldKeyColumns.put(column.getName(), copyEventColumn(column, true, tableHolder));
                // 同时记录一下new
                // key,因为mysql5.6之后出现了minimal模式,after里会没有主键信息,需要在before记录中找
                keyColumns.put(column.getName(), copyEventColumn(column, true, tableHolder));
            } else {
                if (needAllColumns && entry.getHeader().getSourceType() == CanalEntry.Type.ORACLE) {
                    // 针对行记录同步时，针对oracle记录一下非主键的字段，因为update时针对未变更的字段在aftercolume里没有
                    notKeyColumns.put(column.getName(), copyEventColumn(column, isRowMode, tableHolder));
                }
            }
        }
        for (Column column : afterColumns) {
            if (isKey(tableHolder, tableName, column)) {
                // 获取变更后的主键
                keyColumns.put(column.getName(), copyEventColumn(column, true, tableHolder));
            } else if (needAllColumns || entry.getHeader().getSourceType() == CanalEntry.Type.ORACLE || column.getUpdated()) {
                // 在update操作时，oracle和mysql存放变更的非主键值的方式不同,oracle只有变更的字段;
                // mysql会把变更前和变更后的字段都发出来，只需要取有变更的字段.
                // 如果是oracle库，after里一定为对应的变更字段
                boolean isUpdate = true;
                if (entry.getHeader().getSourceType() == CanalEntry.Type.MYSQL) {
                    // mysql的after里部分数据为未变更,oracle里after里为变更字段
                    isUpdate = column.getUpdated();
                }
                // 如果是rowMode，所有字段都为updated
                notKeyColumns.put(column.getName(), copyEventColumn(column, isRowMode || isUpdate, tableHolder));
            }
        }
        if (entry.getHeader().getSourceType() == CanalEntry.Type.ORACLE) {
            // 针对oracle进行特殊处理
            checkUpdateKeyColumns(oldKeyColumns, keyColumns);
        }
    }
    List<EventColumn> keys = new ArrayList<EventColumn>(keyColumns.values());
    List<EventColumn> oldKeys = new ArrayList<EventColumn>(oldKeyColumns.values());
    List<EventColumn> columns = new ArrayList<EventColumn>(notKeyColumns.values());
    Collections.sort(keys, new EventColumnIndexComparable());
    Collections.sort(oldKeys, new EventColumnIndexComparable());
    Collections.sort(columns, new EventColumnIndexComparable());
    if (!keyColumns.isEmpty()) {
        eventData.setKeys(keys);
        if (eventData.getEventType().isUpdate() && !oldKeys.equals(keys)) {
            // update类型，如果存在主键不同,则记录下old
            // keys为变更前的主键
            eventData.setOldKeys(oldKeys);
        }
        eventData.setColumns(columns);
    // } else if (CanalEntry.Type.MYSQL ==
    // entry.getHeader().getSourceType()) {
    // // 只支持mysql无主键同步
    // if (eventType.isUpdate()) {
    // List<EventColumn> oldColumns = new ArrayList<EventColumn>();
    // List<EventColumn> newColumns = new ArrayList<EventColumn>();
    // for (Column column : beforeColumns) {
    // oldColumns.add(copyEventColumn(column, true, tableHolder));
    // }
    //
    // for (Column column : afterColumns) {
    // newColumns.add(copyEventColumn(column, true, tableHolder));
    // }
    // Collections.sort(oldColumns, new EventColumnIndexComparable());
    // Collections.sort(newColumns, new EventColumnIndexComparable());
    // eventData.setOldKeys(oldColumns);// 做为老主键
    // eventData.setKeys(newColumns);// 做为新主键，需要保证新老主键字段数量一致
    // } else {
    // // 针对无主键，等同为所有都是主键进行处理
    // eventData.setKeys(columns);
    // }
    } else {
        throw new SelectException("this rowdata has no pks , entry: " + entry.toString() + " and rowData: " + rowData);
    }
    return eventData;
}