FAQ
I had problems with automatic failover. I noticed that I could not
initialize HA state in zookeeper because the wizard does not deploy the
configuration to the namenodes. So, when the initialization command fails.
I had to manually start the configuration deployment and run the
initialization command. Then, it worked. Is this a bug?

Thanks,
Ben

Search Discussions

  • Vinithra Varadharajan at Dec 20, 2012 at 10:21 pm
    Ben,

    Can you explain the steps that you took and attach the logs (stderr, log4j
    output of FailoverController) of the failed command?

    Deployment of the configuration is not required. Each command uses a
    generated set of configurations - you can see this in
    /var/run/cloudera-scm-agent/process under a directory that will have
    "failover-controller-initialize-znode" in its name. Commands do not use the
    deployed configurations; those are meant for use by clients only. This
    command gets run from the host of the FailoverController, which should be
    the same as your NameNode's host. I suspect that the initialize ZNode
    command failed for other reasons and the logs will help us figure out what
    they are.

    -Vinithra
    On Thu, Dec 20, 2012 at 1:43 PM, Benjamin Kim wrote:

    I had problems with automatic failover. I noticed that I could not
    initialize HA state in zookeeper because the wizard does not deploy the
    configuration to the namenodes. So, when the initialization command fails.
    I had to manually start the configuration deployment and run the
    initialization command. Then, it worked. Is this a bug?

    Thanks,
    Ben
  • Benjamin Kim at Dec 20, 2012 at 10:26 pm
    Vinithra,

    Can you tell me where to look for these logs?

    Thanks,
    Ben
    On Thursday, December 20, 2012 2:21:23 PM UTC-8, Vinithra wrote:

    Ben,

    Can you explain the steps that you took and attach the logs (stderr, log4j
    output of FailoverController) of the failed command?

    Deployment of the configuration is not required. Each command uses a
    generated set of configurations - you can see this in
    /var/run/cloudera-scm-agent/process under a directory that will have
    "failover-controller-initialize-znode" in its name. Commands do not use the
    deployed configurations; those are meant for use by clients only. This
    command gets run from the host of the FailoverController, which should be
    the same as your NameNode's host. I suspect that the initialize ZNode
    command failed for other reasons and the logs will help us figure out what
    they are.

    -Vinithra

    On Thu, Dec 20, 2012 at 1:43 PM, Benjamin Kim <bbui...@gmail.com<javascript:>
    wrote:
    I had problems with automatic failover. I noticed that I could not
    initialize HA state in zookeeper because the wizard does not deploy the
    configuration to the namenodes. So, when the initialization command fails.
    I had to manually start the configuration deployment and run the
    initialization command. Then, it worked. Is this a bug?

    Thanks,
    Ben
  • Vinithra Varadharajan at Dec 21, 2012 at 12:56 am
    Navigate to the FailoverController instance -> Processes -> find the
    initialize znode process -> click on its stderr and log file links.
    On Thu, Dec 20, 2012 at 2:26 PM, Benjamin Kim wrote:

    Vinithra,

    Can you tell me where to look for these logs?

    Thanks,
    Ben

    On Thursday, December 20, 2012 2:21:23 PM UTC-8, Vinithra wrote:

    Ben,

    Can you explain the steps that you took and attach the logs (stderr,
    log4j output of FailoverController) of the failed command?

    Deployment of the configuration is not required. Each command uses a
    generated set of configurations - you can see this in
    /var/run/cloudera-scm-agent/**process under a directory that will have
    "failover-controller-**initialize-znode" in its name. Commands do not
    use the deployed configurations; those are meant for use by clients only.
    This command gets run from the host of the FailoverController, which should
    be the same as your NameNode's host. I suspect that the initialize ZNode
    command failed for other reasons and the logs will help us figure out what
    they are.

    -Vinithra

    On Thu, Dec 20, 2012 at 1:43 PM, Benjamin Kim wrote:

    I had problems with automatic failover. I noticed that I could not
    initialize HA state in zookeeper because the wizard does not deploy the
    configuration to the namenodes. So, when the initialization command fails.
    I had to manually start the configuration deployment and run the
    initialization command. Then, it worked. Is this a bug?

    Thanks,
    Ben

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupscm-users @
categorieshadoop
postedDec 20, '12 at 9:43p
activeDec 21, '12 at 12:56a
posts4
users2
websitecloudera.com
irc#hadoop

People

Translate

site design / logo © 2022 Grokbase