OEM Agent Alert History Graph 2007-11-05 - By Charlotte Hammond
Back
Hi All,
In the off-chance that anyone was interested in the outcome of this....
I traced the OMS sessions to try to find the query being used to populate the
bar graphs (not immediately obvious which query it was but tracing for a very
short instance around viewing the graph limited the choices). Turned out it
was this query:
SELECT :B3 +
(CNT.C*:B8 )/:B1 START_POINT, MAX(SEV.SEVERITY_CODE) SEVERITY_CODE FROM (SELECT
/*+ INDEX(s) */ S.COLLECTION_TIMESTAMP, NVL(S.SEVERITY_DURATION/24, :B2
-S.COLLECTION_TIMESTAMP) SEVERITY_DURATION, S.SEVERITY_CODE FROM MGMT_SEVERITY
S WHERE S.TARGET_GUID = :B7 AND S.METRIC_GUID = :B6 AND S.SEVERITY_CODE IN (:B5
, :B4 ) AND S.COLLECTION_TIMESTAMP+NVL(S.SEVERITY_DURATION/24, :B2
-S.COLLECTION_TIMESTAMP) > :B3 AND S.COLLECTION_TIMESTAMP < :B2 ) SEV,
(SELECT /*+ INDEX(o) */ ROWNUM-1 (See http://NUM-1.ora-code.com) C FROM SYS.OBJ$ O WHERE OBJ# > 0 AND ROWNUM
<= :B1 ) CNT WHERE SEV.COLLECTION_TIMESTAMP >= :B3 + (CNT.C*:B8 )/:B1 -
SEV.SEVERITY_DURATION AND SEV.COLLECTION_TIMESTAMP <= :B3 + ((CNT.C+1)*:B8
)/:B1 GROUP BY :B3 + (CNT.C*:B8 )/:B1
For my single collection error it produced 70 rows of data - at chronological
intervals. However the data was returned in an arbitrary (non-chronological)
order.
Sorry, BAAG guys :-), but I then guessed that this lack of ordering was the
problem, and so set OPTIMIZER_FEATURE_ENABLE = 9.2.0 and flushed the shared
pool. Doing this caused the GROUP BY to also do an implicit ordering and
return the data in chronological order.
Re-freshing the graph in the OEM browser window then redrew the graph exactly
as I expected as a big chunk of red while the target was down, not dozens of
random vertical lines all over the place.
It looks like OEM expects this data to be date-ordered but is relying on GROUP
BY to do it - which won't work with the 10g optimizer.
I've asked a colleague to raise an SR on this - see if we can keep with
OPTIMIZER_FEATURE_ENABLE = 9.2.0 as a workaround, although obviously getting
Oracle to fix the SQL would be better.
Charlotte.
-- --Original
Message-- --
From: oracle-l-bounce@(protected) [mailto:oracle-l-bounce@(protected)] On
Behalf Of Charlotte Hammond
Sent: 01 November 2007 12:16
To: oracle-l@(protected)
Subject: OEM Agent Alert History Graph
Hi All,
I'm using OEM 10.2.0.2 and am having a
problem with a couple of agents.
If, on the Agent's main page, I click on
"Alert History" I see the Metric "Count of targets not uploading
data" I can see, in bar graph format, lots of red vertical lines indicating
critical state in the last 24 hour. The
pattern is "random" but sometimes several an hour and sometimes 3 or
more hours go by without any.
However, if I click on the bar graph and
get the same information in tabular format there are much, much fewer incidents
(maybe 1 or 2 per day). Given the
collection schedule is 240 minutes, I can't see how this metric is appearing in
the bar graph several times an hour. The
tabular data looks much more as I'd expect it.
Furthermore, on one agent, I think I've
cured the problem with the uploads and I see no alerts either in the bar graph
(view data: last 24 hours) or tabular form, yet when I look at the bar graph
(view data: last 7 days) I do still see red error lines which appear *within*
the last 24 hours.
Therefore, my question is, where is the
data coming from in the repository to draw the bar graph? If I could see the
raw data I'd have a better
idea whether I can trust the bar graph or not.
__ ____ ____ ____ ____ ____ ____ ____ ____ ____ __
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
--
http://www.freelists.org/webpage/oracle-l
|
|
