Re: odd estat numbers

On 14 Nov 2002 16:50:38 -0800, Mark.Powell_at_eds.com (Mark D Powell) wrote:

>spamdump_at_nospam.noway.nohow (Ed Stevens) wrote in message news:<3dd3cbe3.56614988_at_ausnews.austin.ibm.com>...
>> Platform: Oracle EE 8.1.7 on Win2k
>>
>> Feeding a series of estat reports to Oraperf, and started getting back replies
>> that 'timed statistics might not be turned on'. Well, I knew that wasn't the
>> case, so I started comparing and estat report that Oraperf liked vs. one that it
>> didnt'. I don't know if this caused the hiccup, but I found this in the 'bad'
>> report:
>>
>> SVRMGR> Rem System wide wait events for non-background processes (PMON,
>> <snip>
>> Event Name Count Total Time Avg Time
>> -------------------------------- ------------- ------------- -------------
>> SQL*Net message from client 68 0 0
>> SQL*Net message to client 68 0 0
>> control file sequential read 15 0 0
>> db file sequential read 14 0 0
>> file open 15 0 0
>> log file sync 1 0 0
>> refresh controlfile command 4 0 0
>> 7 rows selected.
>> SVRMGR>
>> SVRMGR>
>> SVRMGR> Rem System wide wait events for background processes (PMON, SMON, etc)
>> <snip>
>> Event Name Count Total Time Avg Time
>> -------------------------------- ------------- ------------- -------------
>> control file parallel write 291 0 0
>> control file sequential read 12 0 0
>> db file parallel write 1 0 0
>> log file parallel write 1 0 0
>> pmon timer 292 0 0
>> rdbms ipc message 890 0 0
>> smon timer 3 0 0
>> 7 rows selected.
>> SVRMGR>
>>
>>
>>
>> Notice that in spite of their being counts, the reported times are zero. I
>> might could understand that for low-count items like log_file_sync, the total
>> time might have been below the millisecond threshold, but ALL of them?
>
>Ed, did you turn timed statistics on at the database level?
>(timed_statistics = true in init.ora)
>
>HTH -- Mark D Powell --

Uh, how else *would* it be turned on? ;-)

Actually, at this point I have to quote one of my favorite SNL characters: "Never mind." It turns out that timed_stats apparently was NOT turned on. I assumed it was (I know, I know) and since some other areas of the report were showing reasonable numbers I thought that was confirmed. But when I went back a little more carefully and looked at the part of the report that showed non-default init parms (and the default for timed_stats is "false") I found that timed_stats was NOT listed . . . .

I take a 15-minute report daily, and it was only the older reports (I was running the entire collection thru Oraperf) that showed the problem, so stats got turned on somewhere, but I dont' recall doing it. But then, I don't recall what I had for breakfast yesterday . . .

mea culpa.

--
Ed Stevens
(Opinions expressed do not necessarily represent those of my employer.)