-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/41133/#review109588
-----------------------------------------------------------



ambari-server/src/main/resources/common-services/KNOX/0.5.0.2.2/package/scripts/params_linux.py
 (line 43)
<https://reviews.apache.org/r/41133/#comment169163>

    This is the version that the comment should be on during an RU/EU, why are 
we changing the variable if it's None to the current value pointed by 
/usr/hdp/current/knox-server?
    
    There's another parameter for the stack, such as 2.2 or 2.3, that is used 
during normal operation.



ambari-server/src/main/resources/common-services/SPARK/1.2.0.2.2/package/scripts/params.py
 (line 152)
<https://reviews.apache.org/r/41133/#comment169164>

    Same comments above regarding version.
    
    stack_version_unformatted = str(config['hostLevelParams']['stack_version'])
    hdp_stack_version = format_hdp_stack_version(stack_version_unformatted)
    
    # New Cluster Stack Version that is defined during the RESTART of a Stack 
Upgrade
    version = default("/commandParams/version", None)


- Alejandro Fernandez


On Dec. 9, 2015, 2:19 p.m., Oliver Szabo wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/41133/
> -----------------------------------------------------------
> 
> (Updated Dec. 9, 2015, 2:19 p.m.)
> 
> 
> Review request for Ambari, Alejandro Fernandez and Sebastian Toader.
> 
> 
> Bugs: AMBARI-14270
>     https://issues.apache.org/jira/browse/AMBARI-14270
> 
> 
> Repository: ambari
> 
> 
> Description
> -------
> 
> HiveServer2 and SparkThriftserver uses 10000 port by default. Ambari 
> redefines this port (10015) in case of commandParams/version exists (based on 
> spark-hive-site-override.xml).
> During blueprint provisioning commandParams/version not accesible (when all 
> of the install/start task sent with predefined configurations), because 
> Ambari needs at least 1 installed service package to determine the version.
> 
> On start, service packages already installed, so on that phase 
> /usr/bin/hdp-select is usable. -> I used that to gather the hdp stack 
> version, when version is not available. (did the same with KNOX too)
> 
> Notice that: With Blueprints, role command order applied only on the hosts, 
> not on the cluster, so SparkThriftServer wait for HiveMetastore starts only 
> on the same hosts. That means its possible: other SparkThriftServer cannot be 
> started (but can be restarted after the installation)
> 
> Also I added 3 new properties (from hive-site) to SparkThriftServer: 
> 'hive.metastore.connect.retries', 
> 'hive.metastore.client.connect.retry.delay', 
> 'hive.metastore.client.socket.timeout'
> - default delay is too low, now it set as the same as in hive
> 
> 
> Diffs
> -----
> 
>   
> ambari-server/src/main/resources/common-services/KNOX/0.5.0.2.2/package/scripts/params_linux.py
>  f26874f 
>   
> ambari-server/src/main/resources/common-services/SPARK/1.2.0.2.2/package/scripts/params.py
>  18a0052 
> 
> Diff: https://reviews.apache.org/r/41133/diff/
> 
> 
> Testing
> -------
> 
> Total run:814
> Total errors:0
> Total failures:0
> OK
> INFO: AMBARI_SERVER_LIB is not set, using default /usr/lib/ambari-server
> INFO: Return code from stack upgrade command, retcode = 0
> StackAdvisor implementation for stack HDP1, version 2.0.6 was not found
> Returning DefaultStackAdvisor implementation
> StackAdvisor implementation for stack XYZ, version 1.0.0 was loaded
> StackAdvisor implementation for stack XYZ, version 1.0.1 was loaded
> Returning XYZ101StackAdvisor implementation
> [INFO] 
> ------------------------------------------------------------------------
> [INFO] BUILD SUCCESS
> [INFO] 
> ------------------------------------------------------------------------
> [INFO] Total time: 1:02:05.981s
> [INFO] Finished at: Wed Dec 09 14:11:16 CET 2015
> [INFO] Final Memory: 38M/938M
> [INFO] 
> ------------------------------------------------------------------------
> 
> 
> Thanks,
> 
> Oliver Szabo
> 
>

Reply via email to