Ed Stevens <nospam_at_noway.nohow> wrote in
news:i030jvgf88i73rvgn7ar8ca88n7alfiooc_at_4ax.com:
> On 5 Aug 2003 13:57:01 GMT, Chuck <chuckh_at_softhome.net> wrote:
>
>>Ed Stevens <nospam_at_noway.nohow> wrote in
>>news:no9viv8in76uu23boq6bcb6ma55ld3lut0_at_4ax.com:
>>
>>> On 4 Aug 2003 19:34:35 GMT, Chuck <chuckh_at_softhome.net> wrote:
>>>
>>>>I am trying to set up a custom event in OEM 9.2. It looks like all
>>>>you can return in a custom test is a single value (one row, one
>>>>column). I want to return a value see if the event needs to be
>>>>triggered and a more descriptive report if it is triggered. Knowing
>>>>that a problem exists is almost useless unless you know where the
>>>>problem is.
>>>>
>>>>For example I want to see if there are any segments that can
>>>>allocate < 2 extents before the tablespace is out of space. I want a
>>>>critical alert if 0 extents can be allocated and a warning if 1 can
>>>>be. The problem is I can only return one row with one column to the
>>>>event processor. When the event triggers the associated message is
>>>>simply "current value: n" where n is 0 or 1. I want it to list all
>>>>segments that are in danger of not allocating the next extent and
>>>>how many extents they can allocate. Can this be done?
>>>
>>> Why not just use the stock "tablespace full" event?
>>
>>Because that is based on percent full and the tablespaces in question
>>are so large they will always be at 99 percent full - even when
>>there's enough room for the tables to grow for another month. The
>>tablespaces will always be in warning or alarm state which is almost
>>as bad as not monitoring them at all. I'm sure at some point the
>>alarms will start being ignored.
>>
> Ah, "I see", said the blind man.
>
> With that restriction, I still have a gut feel that there is an easier
> way to solve this, but off hand, can't think of any.
>
> If I understand (that is sometimes a very big "if") the standard
> tablespace full alert would work if it could be tuned down to
> something like "alert when 99.9995 % full, alarm when 99.9997 % full"
> Not that it can be that fine, but if it could . . . . am I right?
>
> <snip>
>
I probably still wouldn't use it. I find it more useful to know when a
segment is going to fail to extend and alert X days in advance. With a %
full based solution, the thresholds would constantly need to be updated. If
I know the growth rate and it's fairly consistent then I can predict that
pretty accurately.
Received on Wed Aug 06 2003 - 10:17:06 CDT
