Page 1 of 1
Post upgrade steps - 5.15 to 6.12
Posted: Sat Sep 07, 2013 1:34 pm
by Timbo
Hi,
Are there any post upgrade steps after migrating from 5.15 ==> 6.0 ==> 6.1?
Looking through the AdminGuide I came across the Index Tool section. Do I need to run these utilities after an upgrade?
Reindex
CheckCRC
Checksort
When/how often should they be run if not after an upgrade?
Thanks
Tim
Re: Post upgrade steps - 5.15 to 6.12
Posted: Mon Sep 09, 2013 2:15 pm
by JimKnicely
Hello,
When you go from 5.15 to 6.0 you will need to re-index!
Check out this post:
viewtopic.php?f=11&t=385&hilit=reindex
Re: Post upgrade steps - 5.15 to 6.12
Posted: Mon Sep 09, 2013 3:36 pm
by Timbo
Hi,
Thanks for the reply, just to confirm:-
- This only needs to be run once after the upgrade?
- I missed this step when going from 5.15 to 6.0, is it ok to run it now on 6.12 with the cluster down?
Are there any recomendations for running CheckCRC and Checksort ?
Thanks
Tim
Re: Post upgrade steps - 5.15 to 6.12
Posted: Mon Sep 09, 2013 4:06 pm
by JimKnicely
Hi,
Yes, you only need to run the re-index once. Any new ROS containers that your database is creating is already using the new index format.
Note that you need to run the re-index tool on each node while the cluster is down. You can run the tool in parallel to save time.
The CheckCRC and Checksort can be run anytime after re-indexing. I ran them right away from the command line on each node since the cluster was down anyway. Although you can run them while the cluster is up later.
The first time you run the CheckCRC tool it'll calculate a a cyclic redundancy check (CRC) on each block in your storage. You can then run it again later at any time to check for data corruption.
The Checksort can be run any time you suspect a ROS container data is not sorted the same as a projection's sort order.
Re: Post upgrade steps - 5.15 to 6.12
Posted: Tue Sep 10, 2013 1:07 pm
by Timbo
Hi Jim,
Thanks, have run the Reindex through ok.
How often should the CheckCRC run, once a week?
Regards
Tim
Re: Post upgrade steps - 5.15 to 6.12
Posted: Tue Sep 10, 2013 1:52 pm
by JimKnicely
Hi Timbo,
I've been working with Vertica 6.0 for a while and have yet to see any corruption <<knock on wood>>
I believe that by default the CheckCRC option is always running...
Code: Select all
dbadmin=> SELECT parameter_name, description, default_value, current_value FROM configuration_parameters WHERE parameter_name = 'CheckCRCs';
parameter_name | description | default_value | current_value
----------------+---------------------------------------------------------------------------------------------+---------------+---------------
CheckCRCs | Check data integrity using CRCs (should be enabled unless performance is adversely impacted | 1 | 1
(1 row)
If CRC errors occur while fetching data you'll see errors in your vertica.log file like "CRC Check Failure Details".
You can use an SNMP trap to capture CRC errors via the event manager.