Ck(三) 灵活批量写入大量数据至多张集群表

一、前言

今天有一个批量写入多张集群表数据的任务。要求集群每个节点都要写入一定数据。于是我多花了几个小时，写了一个灵活的批量写入代码。以后再有类似任务可以直接拿来用。经测试是没问题的，但是细节可能写的不好。有空再改。

主要解决问题：Clickhouse批量写入多个节点数据。

二、开始

1、首先是Clickhouse的数据源初始化。正常使用是考虑使用BalancedClickhouseDataSource，但是我想代码控制写入数据量的均匀，所以使用的ClickHouseDataSource。

public class ClickhouseDataSource {

    private volatile static List<DataSource> sources = null;

    public static List<DataSource> getDataSource() {
        if (sources == null) {
            synchronized (ClickhouseDataSource.class) {
                if (sources == null) {
                    List<DataSource> result = Lists.newArrayList();
                    List<Map<String, Object>> collect = ConfigUtils.getConfig().getList("store.clickhouse.connect").stream().map(configValue -> (Map<String, Object>) configValue.unwrapped())
                            .collect(Collectors.toList());
                    List<String> urlList = collect.stream().map(node -> (String) node.get("connection-url")).collect(Collectors.toList());
                    Map<String, Object> unwrapped = (Map<String, Object>) ConfigUtils.getConfig().getValue("store.clickhouse.param").unwrapped();
                    ClickHouseProperties ckProperties = new ClickHouseProperties();
                    ckProperties.setMaxBlockSize(80000 * 1000);
                    ckProperties.setMaxMemoryUsage(300000000000L);
                    ckProperties.setMaxTotal(1000);
                    ckProperties.setUseServerTimeZone(false);
                    ckProperties.setUseServerTimeZoneForDates(false);
                    ckProperties.setUseTimeZone((String) unwrapped.get("zone"));
                    ckProperties.setDefaultMaxPerRoute(500);
                    ckProperties.setConnectionTimeout(1500 * 1000);
                    ckProperties.setKeepAliveTimeout(-1);
                    ckProperties.setSocketTimeout(Integer.MAX_VALUE);
                    //ckProperties.setUser((String) unwrapped.get("user"));
                    //ckProperties.setPassword((String) unwrapped.get("password"));
                    sources = Lists.newArrayList();
                    for (String url : urlList) {
                        DataSource ds = new ClickHouseDataSource(url, ckProperties);
                        sources.add(ds);
                    }

                }
            }
        }
        return sources;
    }
}

2、ck 批量写入。将数据转成TSV后写入。ClickhouseDao中通过dbName和tableName查询system.columns来获得列的名称和类型。即知道类型那么即可根据类型来模拟数据

public class BatchStringHttpEntity extends AbstractHttpEntity {

    private List<String> batchRows;
    public static final String NULL = "\\N";

    public BatchStringHttpEntity(List<String> batchRows) {
        this.batchRows = batchRows;
    }

    @Override
    public boolean isStreaming() {
        return false;
    }

    @Override
    public boolean isRepeatable() {
        return true;
    }

    @Override
    public long getContentLength() {
        return -1;
    }

    @Override
    public InputStream getContent() throws IOException, IllegalStateException {
        throw new UnsupportedOperationException();
    }

    @Override
    public void writeTo(OutputStream outputStream) throws IOException {
        for (String row : batchRows) {
            outputStream.write(row.getBytes(StreamUtils.UTF_8));
        }
    }
}

@Slf4j
public class ClickhouseDao {

    private DataSource dataSource;
    public ClickhouseDao(DataSource dataSourc