-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/36592/
-----------------------------------------------------------
Review request for Ambari, Jonathan Hurley, Nate Cole, and Sid Wagle.
Bugs: AMBARI-12455
https://issues.apache.org/jira/browse/AMBARI-12455
Repository: ambari
Description
-------
Support has identified the need to come up with a script to fix any mismatches
in the database, or identify problems during Rolling Upgrade.
This can be a simple Python script that,
- On a newly installed cluster, ensures that there is at least one cluster
version whose state is CURRENT. If not, will advise the user to restart
services.
- If the user has Registered and Installed repos, check that each one has a
unique version and display name. Further, if any are stuck in an INSTALLING
state, will let the user take three potential actions: leave as is, force to
INSTALLED, force to INSTALL_FAILED.
- If the user has Registered and Installed repos, and one cluster_version is
already in an UPGRADING state, perhaps because hdp-select changed the symlinks
and a component was restarted, or the user inadvertently started a manual
upgrade, will allow the user to force it back to INSTALLED.
- If the user in the in the middle of an upgrade, and they want to force one of
the versions are CURRENT, it will update all DB records accordingly, and set
the previously CURRENT version to INSTALLED.
For now, this will support Ambari 2.0.0 and higher, and MySQL, and Postgres.
Diffs
-----
ambari-server/src/main/resources/scripts/ru_magician.py PRE-CREATION
Diff: https://reviews.apache.org/r/36592/diff/
Testing
-------
Tested on Ambari 2.1.0 with Postgres, and Ambari 2.0.1 on MySQL (using local
and external DB), all of the scenarios above.
I still have to do more thorough testing in the cases of Finalize, but this is
a good preliminary patch to start getting some feedback.
I could have used SQLAlchemy as my ORM, but I really wanted a standalone script
that could be a quick v1, so I went for raw SQL that works on both Postgres &
MySQL.
Thanks,
Alejandro Fernandez
