Problems with Hive Metastore and HiveServer2 after cluster restart

to create new topics or reply. | New User Registration

This topic contains 3 replies, has 2 voices, and was last updated by  Larry Liu 2 years, 4 months ago.

  • Creator
  • #19101



    I recently installed HDP 1.2 on RHEL6. This is on an evaluation cluster, with a single node running all the master services and two slave nodes. The installation went well this time, starting all the services and passing all the Smoketests successfully.

    After noticing that the memory usage was approaching 90% on the master node I decided to increase the memory from 8 to 16GB (the master is on a VM). I followed the cluster shutdown instructions at, shut down the host, increased the memory, and then followed the startup procedure in chapter 3.1.

    There were a few problems starting Oozie, Ganglia, and WebHCat, but I was able to resolve those.

    The problem that I can’t fix is that Hive Metastore and HiveServer2 don’t appear to be coming up in Ambari. Ambari shows a red dot next to “Hive/HCat” under the Services tab, and when I hover over the red dot in the Hosts tab it says “The following master components are down: Hive Metastore, HiveServer2″.

    Now I don’t even know if this is accurate. Nagios shows that everything is good, but I also don’t see specific checks for these two services. I’m not sure where Ambari is pulling this information from.

    Running the command

    !connect jdbc:hive2://$hive.server.full.hostname:10000 $HIVE_USER password org.apache.hive.jdbc.HiveDriver

    appears to be successful, and i can list tables and generally have a working connection to Hive. But netstat doesn’t show an open port on 10000, so I’m not 100% sure that this is accurate.

    Port 9083 on the master is listening according to netstat. I believe is the HCatalog port. But the Hive log shows two errors:

    WARNING: org.apache.hadoop.metrics.jvm.EventCounter is deprecated. Please use org.apache.hadoop.log.metrics.EventCounter in all the files.
    org.apache.thrift.transport.TTransportException: Could not create ServerSocket on address


    Exception in thread “main” org.apache.thrift.transport.TTransportException: Could not create ServerSocket on address

    Hive seems to be functioning just fine. I was able to create tables and query them.

    So the main questions I have are
    1. Are HCatalog and HiveServer2 running but detected by Ambari as not running for some reason? Ambari gives me the option to start Hive and HCat, but it doesn’t change the status when I do.
    2. If the problem is not with Ambari, where do I continue troubleshooting?

    All help is very much appreciated.
    2. Wh

Viewing 3 replies - 1 through 3 (of 3 total)

You must be to reply to this topic. | Create Account

  • Author
  • #19288

    Larry Liu

    Hi, Kevin

    Please delete them, stop services. And try to start from Ambari UI.




    Hi Larry,

    Thanks for taking a look. I do see the two PID files, but there are no processes running with the PID numbers inside even though Hive is up and running. What do you suggest I do? I might try to delete them and restart the services.



    Larry Liu

    Hi, Kevin,

    Thanks for trying HDP.

    It looks like you have Ambari installation. It is not recommended to do any start/stop from command line if your installation is Ambari. Starting/stopping services can be done via Ambari UI as well.

    In HDP 1.2.2, Ambari should be able to sync the status of the services with the actual status on the server. If you see this is not happening, there should be something wrong. Ambari depends on pid file to sync up the services status. For hive, the pid files should be at /var/run/hive. You should be able to find 2 pid files: and

    If you couldn’t find the 2 pid files, please check hive.log and hive-server2.log.

    Hope this helps.


Viewing 3 replies - 1 through 3 (of 3 total)
Hortonworks Data Platform
The Hortonworks Data Platform is a 100% open source distribution of Apache Hadoop that is truly enterprise grade having been built, tested and hardened with enterprise rigor.
Get started with Sandbox
Hortonworks Sandbox is a self-contained virtual machine with Apache Hadoop pre-configured alongside a set of hands-on, step-by-step Hadoop tutorials.
Modern Data Architecture
Tackle the challenges of big data. Hadoop integrates with existing EDW, RDBMS and MPP systems to deliver lower cost, higher capacity infrastructure.