[HBASE-7111] hbase zkcli will not start if the zookeeper server chosen to connect to is unavailable

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.94.2
• Fix Version/s: 0.98.0, 0.94.6, 0.95.0
• Component/s: Zookeeper
• Hadoop Flags: Reviewed
• Assignee: Zhou wenjian
• Reporter: Zhou wenjian
• Created: 07/Nov/12 03:08
• Updated: 07/May/13 03:51
• Resolved: 04/Mar/13 21:56

Description

there are 3 zookeeper servers in my cluster.
s1
s2
s3
after killing s3, i found the hbase zkcli will not start again.
it will try to connect to s3 continuely.

/11/07 11:01:01 INFO zookeeper.ClientCnxn: Opening socket connection to server s3
12/11/07 11:01:01 WARN zookeeper.ClientCnxn: Session 0x0 for server null, unexpected error, closing socket connection and attempting reconnect
java.net.ConnectException: Connection refused

from the code

  public String parse(final Configuration c) {
    // Note that we do not simply grab the property
    // HConstants.ZOOKEEPER_QUORUM from the HBaseConfiguration because the
    // user may be using a zoo.cfg file.
    Properties zkProps = ZKConfig.makeZKProps(c);
    String host = null;
    String clientPort = null;
    for (Entry<Object, Object> entry: zkProps.entrySet()) {
      String key = entry.getKey().toString().trim();
      String value = entry.getValue().toString().trim();
      if (key.startsWith("server.") && host == null) {
        String[] parts = value.split(":");
        host = parts[0];
      } else if (key.endsWith("clientPort")) {
        clientPort = value;
      }
      if (host != null && clientPort != null) break;
    }
    return host != null && clientPort != null? host + ":" + clientPort: null;
  }

the code will choose the fixed zookeeper server (here is the unavailable s3), which leads to the script fails