[Tora-develop] Re: Wait event analyzer in TOra.

[Henrik J. <hp...@gl...>]

Frits Hoogland wrote:

>
>-----Original Message-----
>From: Henrik Johnson [mailto:hp...@gl...]
>Sent: vrijdag 21 december 2001 18:51
>To: Frits Hoogland; Tora Mailinglist
>Subject: Re: Wait event analyzer in TOra.
>
>
>Frits Hoogland wrote:
>
>>still love the wait event option.
>>
>>some more remarks:
>>-my collegues use it on windows.
>>it crashes much. I know you do not put much energy in the windows version
>>(as I wouldn't do either), but a more stable version would be preferable as
>>they now can not run it.
>>
>I found a very handy tool for debugging crashes which gives me the line 
>it crashed on when I get the 8 digit hexadecimal number that windows 
>coughs up on a crash. I must however compile TOra with debugging turned 
>on and the only release I've done that with so far is the 1.2.1 release 
>(Since I didn't have a clue how to use it I didn't find a need to have 
>this info). It will be in the 1.3.1 release, then I can start hoarding 
>those crash addresses. What version are your Windows users using btw?
>
>win2k, winnt, win98
>
I actually meant what version of TOra :).

>>-global remarks regarding the server tuning options:
>>still the pointer turns to the clock if a refresh is specified. Thought it
>>was solved by a second connection?
>>al the already open tabs get a refresh. this is only necessary in tabs that
>>display changable data. options can simply not be changed, parameter
>>
>changes
>
>>can more and more be done in higher oracle versions, but when searching for
>>a specific parameter, the screen gets refreshed. I find that annoying.
>>
>There are a few tabs that still may block the connection. These are the 
>overview and the file I/O and the indicators view. The last one is only 
>updated if it is being displayed so that isn't so much of a problem. 
>I'll try to fix the last ones before the real 1.4.0 release (At least 
>except for the indicators).
>
>>with statistics it is very annoying that it is refreshed. When someone
>>actually uses the statistics, they first going to sort it on name, value or
>>delta. it is -really- annoying when it is refreshed then.
>>
>I'll remove the automatic refresh for all views except the ones with 
>charts in them. I've found this annoying myself.
>
>>when above mentioned adaptions are applied, the screens get better workable
>>on one hand, and have to fetch less data what helps performance on the
>>
>other
>
>>does the blocking locks tab work?
>>
>I think so, I tried it here by creating a few locks and they certainly 
>showed up. Let me know if it doesn't.
>
>no. doesn't on at least oracle version 7.3, 8.0.5, 8.1.6, 8.1.7.
>
I just tried this by doing a "delete from {table}" in two separate 
sessions. The second one will lock. Then I go into the blocking locks 
tuning tab and it clearly shows the locked session. Do you have an 
example that I can try that doesn't work?

>>-alert tool
>>My opinion is that the name is not well chosen. Everyone I show this tools
>>thinks it has something to do the oracle alert log. think of names like
>>'messaging channel' or something.
>>
>>The way it works is not that simple. think of auto-naming, with the option
>>to rename. something like oracleuser + machine name + numbering if the
>>combination of the two exists.
>>
>>It can be very handy if two tora users are working on a database and need
>>
>to
>
>>share SQL in a simple way.
>>
>Ok, I can change the default name you listen to from TOra to something 
>like that (Hmm, how do I get hold of the hostname on Windows). The usage 
>you describe above isn't really what this was meant for, but it sounds 
>like a good idea :). The reason it is called Alert Tool is that it uses 
>the DBMS_ALERT package. I'm probably going to add an alert log viewer 
>before 1.4, I'll try to find a better name for this.
>
>the hostname is fetched when looking at the machine column from the
>v$session view.
>what was the inteded meaning of the alerting tool?
>how are you going to read the alert log?
>
First of all the name. How about "Alert Messaging". The DBMS_ALERT 
package can be used as a fast way to pass signals between two database 
applications. Regarding the alert log reading I was thinking about an 
application to make it easier to read the logs, you still need to pass 
the actual logfile to TOra since I don't know any way to read the alert 
log contents from within Oracle.

>>-statistics manager
>>the statistics manager's viewpoint is from tables. indexes can -seperately-
>>be analyzed in oracle. think of a object name column, and a type column.
>>
>Sure.
>
>the view per schema makes sense, an option for all object would be handy,
>not necessary.
>an option to quickly see all analyzed objects and all not analyzed objects
>would be handy also.
>
Ok.

>>-sga trace
>>haven't look much into this, but get much 'ora-942 table or view does not
>>exists  EXPLAIN PLAN SET STATEMENT_ID='Tora 1.... etc.
>>-> if the plan table does not exists, as the user if tora must make it.
>>other tools can create it too (like tkprof) I saw this functionality one in
>>tora, but sometimes It doesn't do it. oh, wel lets look, the 942 is because
>>of the next point:
>>-> the explain plan statement does take the query, but does not put owner
>>name in front of objects in the from clause. so it sometimes can't find the
>>objects and gives the ora942.
>>or 'alter user set current_schema = xxx'
>>
>I had no idea you could do this. Will do.
>
>suggest you try it first. I know the imp/exp utilities use (this is the
>correct syntax:)
>alter session set current_schema = <schema>;
>the schema name must be given without quotes.
>
>very few tools have the capacity to generate a explain plan for each
>statement that comes up in the sga. would be great.
>
>>-> all cursors ('sql text') that begin with 'begin' are anonymous sql
>>blocks. they can not be explained, suggestions for  are a small message, or
>>a search through the package for sql-statements (which can be explained
>>using the exec.plan tab).
>>
>This is a really good idea. Especially since now with the SQL parser 
>I've made for the auto indentation it is pretty easy to do.
>
>>
>>merry xmas,
>>will mail more, but need to eat.
>>
>You to.
>
>/Mauritz
>GlobeCom AB
>
>it would be nice if the wait events can be monitored over a longer period of
>time.
>it's only necessary if it is stated explicitly. it is for monitoring
>batchprocesses. I am not there at night (most of the time).
>
>I can give you help regarding tuning and internal oracle thingies. not about
>plsql.
>
I think this is a good idea as well. It's quite a big change though so 
I'm not making any promises when I might get around to adding this.

/Mauritz
GlobeCom AB