ambari-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jonathan Hurley" <jhur...@hortonworks.com>
Subject Re: Review Request 37499: RU: Downgrade Fails For All Components From Finalize
Date Sun, 16 Aug 2015 23:01:06 GMT


> On Aug. 16, 2015, 1:05 a.m., Alejandro Fernandez wrote:
> > ambari-server/src/main/resources/custom_actions/scripts/ru_set_all.py, line 98
> > <https://reviews.apache.org/r/37499/diff/1/?file=1041017#file1041017line98>
> >
> >     Preferable to use the compare_versions function which is easier to find in the
future than hardcoding HDP-2.2
> 
> Jonathan Hurley wrote:
>     I really don't care about the versions here; I care about the stack. It was simple
to say source = HDP-2.3 and target = HDP 2.2 ... what's the equivalent code using compare_versions?
I'd have to convert the values of `downgrade_from_version` and `version` into normalized versions
first, and then compare? Can you give an example?
> 
> Alejandro Fernandez wrote:
>     We have a lot of logic that compares the stack against values like 2.2 or 2.3, and
it was done because it was easier at the time to keep adding that type of logic. The rest
of the scripts use the compare_version function, so I was suggesting being consistent. I'm
ok with the current patch.

I already patched it with your suggestion :)

Can you please check my logic for correctness on compare_version - The unit tests passed,
but another set of eyes would be good.


- Jonathan


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/37499/#review95535
-----------------------------------------------------------


On Aug. 16, 2015, 10:52 a.m., Jonathan Hurley wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/37499/
> -----------------------------------------------------------
> 
> (Updated Aug. 16, 2015, 10:52 a.m.)
> 
> 
> Review request for Ambari, Alejandro Fernandez and Nate Cole.
> 
> 
> Bugs: AMBARI-12778
>     https://issues.apache.org/jira/browse/AMBARI-12778
> 
> 
> Repository: ambari
> 
> 
> Description
> -------
> 
> When downgrading a rolling upgrade from HDP 2.3 back to HDP 2.2, all components will
fail with a circular symlink reference. 
> 
> The problem is that Ambari changes /etc/component/conf from a directory to a symlink
when upgrading to HDP 2.3 from HDP 2.2. During the downgrade, once the component is put back
in HDP 2.2, the following is observed:
> ```
> [root@c6401 flume]# ls -l /etc/flume
> total 8
> drwxr-xr-x 3 root  root 4096 Aug 14 15:44 2.3.2.0-2586
> lrwxrwxrwx 1 root  root   34 Aug 14 16:14 conf -> /usr/hdp/current/flume-server/conf
> drwxr-xr-x 2 flume root 4096 Aug 14 15:42 conf.backup
> 
> (/usr/hdp/2.2.7.0-2816/flume/)
> [root@c6401 flume]# ls -l /usr/hdp/current/flume-server/conf
> total 16
> drwxr-xr-x 2 root root 4096 Aug 14 15:42 bin
> lrwxrwxrwx 1 root root   15 Aug 14 15:42 conf -> /etc/flume/conf
> drwxr-xr-x 9 root root 4096 Aug 14 15:42 docs
> drwxr-xr-x 2 root root 4096 Aug 14 15:42 lib
> drwxr-xr-x 2 root root 4096 Aug 14 15:42 tools
> ```
> 
> The solution is to reverse the symlink work done when downgrading from HDP 2.3 back to
2.2.
> 
> 
> Diffs
> -----
> 
>   ambari-common/src/main/python/resource_management/libraries/functions/conf_select.py
a676f74 
>   ambari-server/src/main/resources/common-services/FLUME/1.4.0.2.0/package/scripts/flume_handler.py
a305ce7 
>   ambari-server/src/main/resources/common-services/FLUME/1.4.0.2.0/package/scripts/flume_upgrade.py
7743619 
>   ambari-server/src/main/resources/common-services/FLUME/1.4.0.2.0/package/scripts/params_linux.py
a7efbf6 
>   ambari-server/src/main/resources/custom_actions/scripts/ru_set_all.py f868002 
>   ambari-server/src/main/resources/stacks/HDP/2.2/upgrades/upgrade-2.3.xml f6823c8 
>   ambari-server/src/test/python/custom_actions/test_ru_set_all.py 211443b 
> 
> Diff: https://reviews.apache.org/r/37499/diff/
> 
> 
> Testing
> -------
> 
> Verified fix on a cluster where the problem was reproduced.
> 
> mvn clean test
> ----------------------------------------------------------------------
> Total run:789
> Total errors:0
> Total failures:0
> OK
> 
> 
> Thanks,
> 
> Jonathan Hurley
> 
>


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