Re: Grid Control - opinions please
- From: "Jared Still" <jkstill@xxxxxxxxx>
- To: "Jason Heinrich" <jheinrich@xxxxxxxx>
- Date: Thu, 6 Apr 2006 07:44:02 -0700
Comments below:
On 4/6/06, Jason Heinrich <jheinrich@xxxxxxxx> wrote:
>
> That said, things aren't all sunshine and roses. Running in a browser
> does
> limit you to only working on one target at a time, though you could open
> multiple browser tabs to get around this. Certain targets have a bad
> habit
> of going into "Status pending" or "Metric collection error" status for no
> apparent reason; and I'm having the same issue as Jeffrey Beckstrom, where
> GC won't send me notification emails, even though everything appears to be
> setup properly. Finally, I haven't found a way to display a list of
> sessions on a database without using the Diagnostics Pack.
>
>
I can't help myself here, as there are now multiple complaints about the
notifications.
The Perl/shell/SQLPlus/PLSQL daemons and/or scripts run from cron
have for several years been monitoring databases that I am in responsible
for.
(Mostly Perl)
Among the things monitored:
* database uptime. during (configurable per each db) uptime hours, the
oncall DBA is paged if a database is unresponsive. If it happens during
non-critical hours, the page is delayed until a configurable number of
connection attempts have failed in succession. This avoids waking the
DBA up (me) if the problem corrects itself.
* space - email a report every morning with space issues. Objects
appear in the report if they do not have enough space to add
more extents (5 I think). Accounts for autoextend files.
* Windows services. Windows has the nasty habit of sometimes failing
to start a service on bootup. This could be a timeout issue, as there is
a lot happening at bootup. A Perl service on another windows box
checks certain remote services, restarts them if they are down and
pages the DBA.
* response time monitoring. No paging, just charts generated showing
average response times from the database perspective. Based on
Statspack and YAPP.
* alert log monitoring. check for errors in the alert.log. Which errors
to check and which to ignore is configurable via regex. Runs as a
daemon on linux/unix, and a service on Windows.
This has been in use and working with very few problems since 1998
on Unix, 2002 on Windows. It does work better on unix though.
No doubt many DBA's have reliable home grown scripts and processes
for monitoring and managing databases.
A GUI is no substitute for knowing how the database works.
The scripting approach has an advantage that OEM/GC or any other
tool has: you are not dependent on a vendo (Oracle or anyone else)
to fix or support their tool so you can do your job.
You can create tools to do exactly what you need. If your needs change
you can modify them. When they break, you can fix them.
If you made it this far, thanks for reading all this.
And perhaps you can understand the apprehension felt by many when attempting
to use some of the 'time saving' tools, and the reluctance to give up tried
and true methods.
--
Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist
Other related posts:
