Analyze Workload operation has failed on Database

Pietro_La_Torre

Hi!
I am using Vertica 6.1.3 and MC.
When I use Analyze Workload from MC I get the following Error on Message Center:

Analyze Workload operation has failed on Database! - Could not connect to Database.

BUT :

• actually MC is connected to the DB 
• If I analyze current sessions I can see "ANALYZE_WORKLOAD()"
• The tuning recommendations are refreshed , so the workload is executed correctly.

Why do I get this message? Is there a way to solve this issue?

thank you,
Pietro.

Daniel_Leybovic

Hi!

...hm...may be a lot of reasons.
WLA is a heavy operation so it can be some timeout trigger due long connection with no response and it can be anywhere in configurations: TomCat?(what is a servlet container?), in server(system with Vertica), in client(MC/system).

Can you check when connection closed/aborted?
* you can try via session profiling(but you need to enable it first)
* from table sessions extract a session_id and in logs (vertica.log) to check when it aborted.
(actually check all session, with grep its very easy while session_id is unique)
* im not familiar with MC, but I believe MC also has logs


[UPDATE]

* WLA can run in background while connection is lost, so task is finished, but you don't know about it (so may be it a feature and just no message suppression )
* if you get it constantly/frequently - check interval/delta when query/session aborted. Does interval around some value(+/-1sec)? If so, check environment for timeout trigger.

R_McKeon

Workload Analyze kicks off every day, once a day, in the Management Console at the same time.

Sometimes, this task fails, for no reason and the following is reported in the mconsole.log:

SUMMARY:Analyze Workload operation has failed on Database
Time of occurance:27 Jan 2014 10:01:33
Number of occurances of this message:1 (Last occurance at: 27 Jan 2014 10:01:33)
Description:[''Could not connect to database'']

The connection to the database does work, since manually starting the Analyze Workload from the Management Console should work.

Solution:

To get rid of the errors, remove the cluster from the Management Console display and then re-import.

This usually corrects the problem.