ambari-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alejandro Fernandez (JIRA)" <>
Subject [jira] [Commented] (AMBARI-10055) RU failed on RESTART OOZIE/OOZIE_SERVER because hadoop-lzo should only be copied if lzo compression is enabled
Date Fri, 13 Mar 2015 01:14:38 GMT


Alejandro Fernandez commented on AMBARI-10055:

I verified the patch works.
* Enabled lzo by following
* Ensured that was still able to restart Oozie server
* Took a snapshot of the VM
  ** Installed new version HDP, which correctly downloaded the lzo jar to /usr/hdp/
Was then able to perform a successful RU.
  ** Reverted the LZO configs, and installed new version HDP, which did not download
the lzo jar. Was then able to perform a successful RU.

Basically, the Oozie Server RU will only require the hadoop lzo to exist if lzo is enabled.

There's one corner case in which the RU can still fail because of this: if a user registers
a version and installs it (i.e., distributes the bits to all of the hosts. Note that at this
point there was no need to download lzo jar), and then enables lzo compression, and begins
the RU, then Oozie Server restart will fail because the hadoop-lzo jar was never downloaded.
There are several ways to fix this corner case:
1. Force redistribution of bits for a version that is already INSTALLED. This is more general,
and can help in other scenarios, so I think it's preferable. 
2. Detect what changes to the configs should transition INSTALLED versions as OUT_OF_SYNC;
this is difficult to maintain.
3. Have Oozie Server RU Restart download the package; this is not desirable because the RU
now has to do more work that could have been done up-front.

> RU failed on RESTART OOZIE/OOZIE_SERVER because hadoop-lzo should only be copied if lzo
compression is enabled 
> ---------------------------------------------------------------------------------------------------------------
>                 Key: AMBARI-10055
>                 URL:
>             Project: Ambari
>          Issue Type: Bug
>          Components: ambari-server
>    Affects Versions: 2.0.0
>            Reporter: Alejandro Fernandez
>            Assignee: Alejandro Fernandez
>              Labels: rolling_upgrade
>             Fix For: 2.0.0
>         Attachments: AMBARI-10055.patch
> STR:
> Deploy a cluster with Oozie
> Confirm that lzo compression is not enabled (
> Comment a Rolling Upgrade, which will fail in Oozie because it will not find the hadoop-lzo
> The jar should only be copied if lzo is enabled.
> {code}
> 2015-03-11 21:32:49,019 - u"Execute['hdp-select set oozie-server']" {}
> 2015-03-11 21:32:49,064 - Restoring Oozie configuration directory after upgrade...
> 2015-03-11 21:32:49,065 - Extracting /tmp/oozie-upgrade-backup/oozie-conf-backup.tar
to /etc/oozie/conf
> 2015-03-11 21:32:49,179 - Error while executing command 'restart':
> Traceback (most recent call last):
>   File "/usr/lib/python2.6/site-packages/resource_management/libraries/script/",
line 214, in execute
>     method(env)
>   File "/usr/lib/python2.6/site-packages/resource_management/libraries/script/",
line 365, in restart
>     self.pre_rolling_restart(env)
>   File "/var/lib/ambari-agent/cache/common-services/OOZIE/",
line 165, in pre_rolling_restart
>     oozie_server_upgrade.prepare_libext_directory()
>   File "/var/lib/ambari-agent/cache/common-services/OOZIE/",
line 141, in prepare_libext_directory
>     hadoop_client_new_lib_dir, hadoop_lzo_pattern))
> Fail: There are no files at /usr/hdp/ matching hadoop-lzo*.jar
> 2015-03-11 21:32:49,235 - Command: /usr/bin/hdp-select status oozie-server > /tmp/tmpVfVrGL
> Output: oozie-server -
> {code}

This message was sent by Atlassian JIRA

View raw message