Monitoring database start, stop, create, drop database

DieterC

Hi forum,

a Vertica customer has to implement monitoring for these V events: start, stop, create, drop database

Their organization requires files on these activities to be created an written to a directory where a Splunk client grabs them for transfer.

Where can such info be found in a V cluster please? It is not the dc tables as they are DB specific. Also these events can come from admintools or MC.

Is there a common info source on those events please?
Thank you

Dieter

SruthiA

@DieterC : That information can be found in vertica.log as in when the database was stopped, started etc... the path is dbname/v_dbname_nodexxx_catalog/vertica.log.

mosheg

In addition, the following info can be useful in case one kills the Vertica process without evidence in the vertica.log or
Drop the database, or recreate it with the same name, logrotate or replace vertica.log

1. Spread starts are written to file: /Your_catalog_path/Your_db_name/dbLog
2. Each time Vertica starts or stopped, the following file will change with a new process ID: /Your_catalog_path/DB_name/v_DbName_NodeName/vertica.pid
   One can monitor the file like so:
   nohup tail -F /Your_catalog_path/DB_name/v_DbName_NodeName/vertica.pid > /tmp/check_pid.log 2>&1 &
   tail -F keep trying to open the file if it is inaccessible while Vertica is down.
3. Each time one will drop the database, this directory will be removed: /Your_catalog_path/DB_name
   One can monitor if this directory is deleted with something like: inotifywait -m /Your_catalog_path/DB_name >> /tmp/DB_stat.log

DieterC

Thank you Moshe and Sruthi,

• I do assume vertica.log processing needs to be done for all nodes in the cluster as these are not synchronized?
• could admintools.conf or admintools.log be a source of info? Are these synchronized over the nodes?
• If I had a wish I would want metadata at the cluster level (a level higher than the database)
  Something like the v_catalog schema on a cluster level. Is there a thing close to this?

Thank you
Dieter

mosheg

Yes, any processing needs to be done on all nodes.
/opt/vertica/log/adminTools.log include activity logs made by admintools
But it doesn't cover activities from MC nor SQL.
I think MC does monitor Vertica in a cluster level.
IMHO it is better to use OS tools to monitor Vertica because from OS level Vertica super user like dbadmin or a root user can kill Vertica or the MC processes, and then Vertica cannot log that event.