Question

我使用CQ v5.16.11（带有openjdk 11）来保存每天滚动周期的数据。该过程从星期日到星期五不间断运行，因此我每周有5个cq4文件。我将流程运行了1.5周，共有8个文件（第1周3个文件，第2周5个文件）。

所以我拥有的文件是：

20181003.cq4 cycle=17807,
20181004.cq4 cycle=17808,
20181005.cq4 cycle=17809,
20181007.cq4 cycle=17811,
20181008.cq4 cycle=17812,
20181009.cq4 cycle=17813,
20181010.cq4 cycle=17814,
20181011.cq4 cycle=17815,

请注意缺少20181006.cq4的文件（周期= 17810），因为该进程不在周六运行。

我使用以下代码读取数据：

tailer.toEnd();
lastTailerIndex = tailer.index();
tailer.toStart();

while (tailer.index() <= lastTailerIndex) {
    // read data
    if (tailer.readBytes(data) {
     /// do something with data bytes
    }
    if (tailer.index() == lastTailerIndex) {
        break;
    }
}

这会正确读取第一周数据，但不会读取第二周数据，因为它不会自动滚动到下一个周期。

知道为什么会这样或如何解决吗？

问题类似于旧版本中的问题

日志：

2018-10-12 12:41:15,784 DEBUG [main] net.openhft.chronicle.bytes.MappedFile - Allocation of 0 chunk in /site/data/metadata.cq4t took 19.237 ms.
2018-10-12 12:41:15,876 DEBUG [main] net.openhft.chronicle.bytes.MappedFile - Allocation of 0 chunk in /site/data/20181011.cq4 took 0.063 ms.
2018-10-12 12:41:15,881 DEBUG [main] net.openhft.chronicle.queue.impl.single.PretoucherState - /site/data/20181011.cq4 - Reset pretoucher to pos 4835096 as the underlying MappedBytes changed.
2018-10-12 12:41:15,887 DEBUG [main] net.openhft.chronicle.bytes.MappedFile - Allocation of 0 chunk in /site/data/20181003.cq4 took 0.065 ms.
2018-10-12 12:41:15,995 DEBUG [main] net.openhft.chronicle.bytes.MappedFile - Allocation of 0 chunk in /site/data/20181011.cq4 took 0.082 ms.
2018-10-12 12:41:15,996 DEBUG [main] net.openhft.chronicle.queue.impl.single.SingleChronicleQueueBuilder - File released /site/data/20181003.cq4
2018-10-12 12:41:15,997 DEBUG [main] net.openhft.chronicle.queue.impl.single.SingleChronicleQueueBuilder - File released /site/data/20181011.cq4
2018-10-12 12:41:16,418 DEBUG [main] net.openhft.chronicle.bytes.MappedFile - Allocation of 0 chunk in /site/data/20181004.cq4 took 0.112 ms.
2018-10-12 12:41:16,418 DEBUG [main] net.openhft.chronicle.queue.impl.single.SingleChronicleQueueBuilder - File released /site/data/20181003.cq4
2018-10-12 12:41:16,813 DEBUG [main] net.openhft.chronicle.bytes.MappedFile - Allocation of 0 chunk in /site/data/20181005.cq4 took 0.084 ms.
2018-10-12 12:41:16,813 DEBUG [main] net.openhft.chronicle.queue.impl.single.SingleChronicleQueueBuilder - File released /site/data/20181004.cq4

[编辑1]： 在上一个周末发生了同样的事情，即，按预期，10月13日没有新文件。现在，我有10月7日至10月15日的文件（缺少10月13日的文件）。如果我执行tailer.toStart(); while(tailer.readBytes() { ...}，则它将仅读取10月7日至10月12日的文件，而不会读取10月14日和15日。

[编辑2]：复制了以下问题 Chronicle-Queue/issues/537

设置/库：：jvm openjdk 11，Ubuntu 16.04，openhft.affinity / 3.1.9，编年史地图/3.16.0，编年史队列/5.16.11，记录字节/1.16.23，记录核心/1.16.20，编年史钢丝/1.16.16，编年史线程/1.16.3，jna/4.4.0
步骤：
- 启动WriterProcess-完成它。
- 启动ReaderProcess-请参阅5条打印语句。
- 停止ReaderProcess
- 等待一段时间-10分钟。
- 再次启动WriterProcess-让它完成或继续运行此过程。
- 启动ReaderProcess-它仅打印前5个打印语句，此后不打印任何内容。即使WriterProcess是正在运行/写入以在此过程中将尾部放入队列中不动向前。

public class WriterProcess {
        public static void main(String[] args) throws InterruptedException {
            final String dir = "/tmp/demo/";
            final LocalTime localTime = LocalTime.of(17, 0);
            final ZoneId zoneID = ZoneId.of("America/New_York");
            final ScheduledExecutorService scheduledExecutorService = Executors.newScheduledThreadPool(2);
            final SingleChronicleQueue queue  = SingleChronicleQueueBuilder.binary(dir)
                    .blockSize((long) Math.pow(2, 23))
                    .rollCycle(RollCycles.MINUTELY)
                    .rollTime(localTime, zoneID)
                    .build();
            final ExcerptAppender appender = queue.acquireAppender();
            // pre touch
            scheduledExecutorService.scheduleAtFixedRate(appender::pretouch,0,30, TimeUnit.SECONDS);
            // write data
            System.out.println("writing data ...");
            writeData(appender, 5);
            // close queue
            System.out.println("shutting down now ...");
            queue.close();
            scheduledExecutorService.shutdown();
            scheduledExecutorService.awaitTermination(1, TimeUnit.SECONDS);
        }
        public static void writeData(ExcerptAppender appender, int count) {
            int ctr = 0;
            String dateStr;
            Date date = new Date();
            while (true) {
                dateStr = date.toString();
                appender.writeText("["+ctr+"] Written " + dateStr);
                System.out.println("["+ctr+"] Written " + dateStr);
                ctr++;
                if (ctr >= count) {
                    break;
                }
                try {
                    Thread.sleep(65_000);
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
            }
        }
    }
    public class ReaderProcess {
        public static void main(String[] args) {
            final String dir = "/tmp/demo/";
            final LocalTime localTime = LocalTime.of(17, 0);
            final ZoneId zoneID = ZoneId.of("America/New_York");
            final SingleChronicleQueue queue  = SingleChronicleQueueBuilder.binary(dir)
                    .blockSize((long) Math.pow(2, 23))
                    .rollCycle(RollCycles.MINUTELY)
                    .rollTime(localTime, zoneID)
                    .build();
            final ExcerptTailer tailer = queue.createTailer();
            tailer.toStart();
            // read data
            System.out.println("reading data ...");
            readData(tailer, 25);
            // close
            System.out.println("shutting down now ...");
            queue.close();
        }
        public static void readData(ExcerptTailer tailer, int count) {
            int ctr = 0;
            Bytes data = Bytes.allocateDirect(new byte[500]);
            while (true) {
                if (tailer.readBytes(data)) {
                    System.out.println("["+ctr+"] Read {"+ data + "}");
                    ctr++;
                    if (ctr >= count) {
                        break;
                    }
                }
            }
        }
    }

Answer 1

我写了一个稍微更简单的版本，该版本适用于编年史2.17及其使用的版本。我所做的最大更改是在读取之前清除了字节data，否则它只会追加以免覆盖任何内容。

import net.openhft.chronicle.bytes.Bytes;
import net.openhft.chronicle.core.OS;
import net.openhft.chronicle.queue.ExcerptAppender;
import net.openhft.chronicle.queue.ExcerptTailer;
import net.openhft.chronicle.queue.RollCycles;
import net.openhft.chronicle.queue.impl.single.SingleChronicleQueue;
import net.openhft.chronicle.queue.impl.single.SingleChronicleQueueBuilder;

import java.time.LocalDateTime;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.TimeUnit;

public class WriterProcess {
    static final String dir = OS.TMP + "/demo-" + System.nanoTime() + "/";

    public static void main(String[] args) throws InterruptedException {
        final ScheduledExecutorService scheduledExecutorService = Executors.newScheduledThreadPool(2);
        final SingleChronicleQueue queue = SingleChronicleQueueBuilder.binary(dir)
                .testBlockSize()
                .rollCycle(RollCycles.TEST_SECONDLY)
                .build();
        final ExcerptAppender appender = queue.acquireAppender();
        // pre touch
        scheduledExecutorService.scheduleAtFixedRate(appender::pretouch, 3, 30, TimeUnit.SECONDS);
        new Thread(ReaderProcess::main).start();
        // write data
        System.out.println("writing data ...");
        writeData(appender, 100);
        // close queue
        System.out.println("shutting down now ...");
        queue.close();
        scheduledExecutorService.shutdown();
        scheduledExecutorService.awaitTermination(1, TimeUnit.SECONDS);
    }

    public static void writeData(ExcerptAppender appender, int count) {
        int ctr = 0;
        while (true) {
            LocalDateTime date = LocalDateTime.now();
            appender.writeText("[" + ctr + "] Written " + date);
            System.out.println("[" + ctr + "] Written " + date);
            ctr++;
            if (ctr >= count) {
                break;
            }
            try {
                Thread.sleep(2_200);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        }
    }
}

class ReaderProcess {
    public static void main(String... args) {
        final String dir = WriterProcess.dir;
        final SingleChronicleQueue queue = SingleChronicleQueueBuilder.binary(dir)
                .testBlockSize()
                .rollCycle(RollCycles.TEST_SECONDLY)
                .build();
        final ExcerptTailer tailer = queue.createTailer();
        tailer.toStart();
        // read data
        System.out.println("reading data ...");
        readData(tailer, 100);
        // close
        System.out.println("shutting down now ...");
        queue.close();
    }

    public static void readData(ExcerptTailer tailer, int count) {
        int ctr = 0;
        Bytes data = Bytes.allocateDirect(64);
        while (true) {
            data.clear();
            if (tailer.readBytes(data)) {
                System.out.println("[" + ctr + "] Read {" + data + "}");
                ctr++;
                if (ctr >= count) {
                    break;
                }
            }
        }
    }
}

在Chronicle Queue Tailer（v5.16.11）中阅读消息不会自动将索引移动到下一个滚动周期

1 个答案: