ambari-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dmitro Lisnichenko" <>
Subject Re: Review Request 41517: Atlas MetaData server start fails due to missing package while adding Atlas service after Hive service
Date Fri, 18 Dec 2015 14:49:19 GMT

This is an automatically generated e-mail. To reply, visit:

(Updated Dec. 18, 2015, 4:49 p.m.)

Review request for Ambari, Alejandro Fernandez, Dmytro Grinenko, Erik Bergenholtz, Jonathan
Hurley, John Speidel, Nate Cole, Sumit Mohanty, and Tom Beerbower.

Bugs: AMBARI-14420

Repository: ambari

Description (updated)

# Install HDP 2.1 with Ambari 2.2.0 (cluster has Hive service installed, no Atlas)
# Register HDP version and Install packages
# Perform EU to HDP and let it complete
# Add Atlas service to the cluster

Alternative STR:
# Install HDP 2.1 with Ambari 2.2.0 (cluster has Hive service installed, no Atlas)
# Add Atlas service to the cluster

Atlas MetaData server start fails with below error:
Traceback (most recent call last):
File "/var/lib/ambari-agent/cache/common-services/ATLAS/",
line 132, in <module>
File "/usr/lib/python2.6/site-packages/resource_management/libraries/script/", line
219, in execute
File "/var/lib/ambari-agent/cache/common-services/ATLAS/",
line 53, in start
File "/var/lib/ambari-agent/cache/common-services/ATLAS/",
line 40, in configure
File "/var/lib/ambari-agent/cache/common-services/ATLAS/",
line 69, in metadata
content = StaticFile(format('{metadata_home}/server/webapp/atlas.war'))
File "/usr/lib/python2.6/site-packages/resource_management/core/", line 154, in __init__
File "/usr/lib/python2.6/site-packages/resource_management/core/", line 158,
in run
self.run_action(resource, action)
File "/usr/lib/python2.6/site-packages/resource_management/core/", line 121,
in run_action
File "/usr/lib/python2.6/site-packages/resource_management/core/providers/", line
90, in action_create
content = self._get_content()
File "/usr/lib/python2.6/site-packages/resource_management/core/providers/", line
127, in _get_content
return content()
File "/usr/lib/python2.6/site-packages/resource_management/core/", line 51, in __call__
return self.get_content()
File "/usr/lib/python2.6/site-packages/resource_management/core/", line 75, in get_content
raise Fail("{0} Source file {1} is not found".format(repr(self), path))
resource_management.core.exceptions.Fail: StaticFile('/usr/hdp/current/atlas-server/server/webapp/atlas.war')
Source file /usr/hdp/current/atlas-server/server/webapp/atlas.war is not found

Upon inspection found that:
While installing packages in step 2, below package was installed along with hive packages:
2015-12-16 20:05:15,043 - Package['atlas-metadata*-hive-plugin'] {'use_repos': ['HDP-',
'HDP-UTILS-'], 'skip_repos': ['HDP-*']}
2015-12-16 20:05:15,043 - Installing package atlas-metadata*-hive-plugin ('/usr/bin/yum -d
0 -e 0 -y install '--disablerepo=HDP-*' --enablerepo=HDP-,HDP-UTILS- 'atlas-metadata*-hive-plugin'')

Further while adding Atlas service in step 4, the installation of atlas packages are skipped
because the system thinks that ATLAS packages are already present, which leads to the failure
while starting MetaData server
ATLAS MetaData server install log:
2015-12-17 08:03:56,583 - Package['unzip'] {}
2015-12-17 08:03:56,723 - Skipping installation of existing package unzip
2015-12-17 08:03:56,723 - Package['curl'] {}
2015-12-17 08:03:56,747 - Skipping installation of existing package curl
2015-12-17 08:03:56,748 - Package['hdp-select'] {}
2015-12-17 08:03:56,765 - Skipping installation of existing package hdp-select
2015-12-17 08:03:57,047 - Package['atlas-metadata_2_3_*'] {}
2015-12-17 08:03:57,170 - Skipping installation of existing package atlas-metadata_2_3_*
2015-12-17 08:03:57,480 - The hadoop conf dir /usr/hdp/current/hadoop-client/conf exists,
will call conf-select on it for version
2015-12-17 08:03:57,480 - Checking if need to create versioned conf dir /etc/hadoop/
2015-12-17 08:03:57,481 - call['conf-select create-conf-dir --package hadoop --stack-version --conf-version 0'] {'logoutput': False, 'sudo': True, 'quiet': False, 'stderr':
2015-12-17 08:03:57,520 - call returned (1, '/etc/hadoop/ exist already', '')
2015-12-17 08:03:57,521 - checked_call['conf-select set-conf-dir --package hadoop --stack-version --conf-version 0'] {'logoutput': False, 'sudo': True, 'quiet': False}
2015-12-17 08:03:57,561 - checked_call returned (0, '/usr/hdp/ ->

Note -
a. There is a likelihood of this bug not coming if we chose a host for Atlas MetaData server
that is different from any of the Hive hosts, while adding Atlas service.
b. The bug is seen for EU from HDP 2.2 to 2.3 as well

The problem is when Hive is installed, it installs package atlas-metadata_2_3_4_0_3480-hive-plugin-
If after that we add Atlas service to cluster, some packages are not installed.
When we try to install package atlas-metadata_2_3_4_0_3480-,
we are checking presence by using too wide mask 'atlas-metadata_2_3_*', as a result we skip
installing some Atlas packages
So we decided to do this change. It effectively installs more Atlas packages when Hive is

There is a jira to fix all related problems in 2.4.0 release (already committed)


  ambari-server/src/main/resources/stacks/HDP/2.3/services/HIVE/metainfo.xml b6b6bb8 



mvn clean test

checked on live clusters, did not spot any problems


Dmitro Lisnichenko

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message