Bookie could coredump when it is shutting down

[AnonHxy]

BUG REPORT

Describe the bug

When we kill the bookie processor and shut it down, sometimes core dump could happen. Core dump log：

---------------  T H R E A D  ---------------

Current thread (0x00007f14a7128350):  JavaThread "qtp1834755909-1155" [_thread_in_native, id=7358, stack(0x00007f053cbfc000,0x00007f053ccfd000)]

Stack: [0x00007f053cbfc000,0x00007f053ccfd000],  sp=0x00007f053ccfb130,  free space=1020k
Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)
C  [librocksdbjni3874548073448578178.so+0x37b79b]  rocksdb::DBImpl::GetIntProperty(rocksdb::ColumnFamilyHandle*, rocksdb::Slice const&, unsigned long*)+0x2b
C  [librocksdbjni3874548073448578178.so+0x291bb1]  Java_org_rocksdb_RocksDB_getLongProperty+0x171
J 7450  org.rocksdb.RocksDB.getLongProperty(JJLjava/lang/String;I)J (0 bytes) @ 0x00007f1495901a0d [0x00007f1495901920+0x00000000000000ed]
J 9392 c2 org.apache.bookkeeper.bookie.storage.ldb.EntryLocationIndexStats$1.getSample()Ljava/lang/Number; (5 bytes) @ 0x00007f1495344cac [0x00007f1495344be0+0x00000000000000cc]
J 6119 c2 org.apache.bookkeeper.stats.prometheus.SimpleGauge.getSample()Ljava/lang/Number; (10 bytes) @ 0x00007f1495829f84 [0x00007f1495829f20+0x0000000000000064]
J 5356 c2 org.apache.bookkeeper.stats.prometheus.PrometheusMetricsProvider$$Lambda$310+0x0000000800fa6460.accept(Ljava/lang/Object;Ljava/lang/Object;)V (20 bytes) @ 0x00007f14957245f4 [0x0000
7f14957234a0+0x0000000000001154]
J 6418 c2 java.util.concurrent.ConcurrentHashMap.forEach(Ljava/util/function/BiConsumer;)V [email protected] (65 bytes) @ 0x00007f149554f0d8 [0x00007f149554ee40+0x0000000000000298]
J 10308 c2 org.apache.bookkeeper.stats.prometheus.PrometheusMetricsProvider.writeAllMetrics(Ljava/io/Writer;)V (65 bytes) @ 0x00007f149592c3b0 [0x00007f149592c1a0+0x0000000000000210]
j  org.apache.bookkeeper.stats.prometheus.PrometheusServlet.doGet(Ljavax/servlet/http/HttpServletRequest;Ljavax/servlet/http/HttpServletResponse;)V+29

To Reproduce

It's not easy to reprodce the core dump everytime. Howerver we could find that it is related to rocksdb and PrometheusServlet from the core dump log:

1. The PrometheusServlet still services when bookie shutting down half
2. At this time a GET request for Prometheus will call rocksdb::DBImpl::GetIntProperty, while maybe the rocksdb has been closed now. Then core dump happends

So the root casue could be calling rocksdb#getLongProperty after rocksdb has been closed

And we could reproduce the rocksdb core dump from the following code:

 <dependency>
      <groupId>org.apache.bookkeeper</groupId>
      <artifactId>bookkeeper-server</artifactId>
      <version>4.15.4</version>
 </dependency>

public static void main(String[] args) throws Exception {
        RocksDB.loadLibrary();

        try (final ColumnFamilyOptions cfOpts = new ColumnFamilyOptions().optimizeUniversalStyleCompaction()) {
            final List<ColumnFamilyDescriptor> cfDescriptors = Arrays.asList(new ColumnFamilyDescriptor(RocksDB.DEFAULT_COLUMN_FAMILY, cfOpts),
                    new ColumnFamilyDescriptor("test-cf".getBytes(), cfOpts));
            final List<ColumnFamilyHandle> columnFamilyHandleList = new ArrayList<>();
            try (final DBOptions options = new DBOptions().setCreateIfMissing(true).setCreateMissingColumnFamilies(true);
                 final RocksDB db = RocksDB.open(options, "/Users/houxiaoyu/test/testrocksdb", cfDescriptors, columnFamilyHandleList)) {
                try {
                    db.put("key2".getBytes(), "value".getBytes());
                } finally {
                    db.close();
                    // calling `rocksdb#getLongProperty` after rocksdb has been closed
                    db.getLongProperty("rocksdb.estimate-num-keys");
                }
            }
        }
    }