as a matter of fact, both the db and ashmon were running on Windows XP, SP2. I’ll try your suggestion and keep you posted.

Thank you for responding.

Best wishes
Livio

Was this on Windows or Linux?
You can comment out (by starting the line with a pound sign, # )
Blt_ZoomStack $graph
In
ashmon/funcs/ash_display.tcl

Not sure why this would be missing unless you are on LINUX (which I haven’t personally tested). This function isn’t required. It allows zooming in and out on the graph. It’s a function that comes with the BLT install that is part of the ASHMON install in ashmon/lib/blt2.4/graph.tcl

Best
Kyle

I’ve tried to run ashmon against an Oracle 10.2.0 database. I’ve logged in as system and got the following error message:

graphsetup:ERROR: invalid command name “Blt_ZoomStack” !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
ash_init: can’t read “public(ash,groups)”: no such element in array

Connection seemed to be successful. What am I missing?

Thanks and best wishes.

Livio

I’m glad you like S-ASH and ASHMON. I agree to that Oracle ignored the wait interface and my goal with the performance screens in OEM 10g were to expose the wait interface clearly (they call the “top activity” page the “Kyle Page”). I also wanted access to ASH on all versions without having to pay the license and I wanted a graphic interface in a simple executable instead of the slow big complicated web install of OEM 10g, thus I created ASHMON. Unfortunately I am no longer able to work on ASHMON .
All the problems/issues you stated are areas of development in ASHMON that have still been left to do. The main problem I think is that SQL Text is only collected once and hour. It should probably be collected every 15-30 seconds for new SQL. I did try changing the job period for collecting “get_all” to every 15 seconds but it looked like it used too much resources. What I think should be done is taking the get sql text part out of get_all and run it separately every 15-30 seconds.
There is no explain plan on 9i because I use the dbms_xplan.display_cursor which doesn’t exist on 9i. I think get_all collects the explain plan if I recall correctly, so you should be able to display the explain plan from the sash repository.
The output you see when clicking a wait event or session is the shell for having detailed information on a session or wait. If you click on an IO wait event or buffer busy waits you will see some extra output.

Best Wishes
Kyle Hailey

thank you for all the help during the way.

I implemented the S-ASH repository according to described steps.
Separating major blocks into separate files indeed is easier to
understand and implement.

So far I decided to create a separate repository for each database
beeing monitored. That way it is very obvious which database is taking
how much resources on the repository server and I like the separation.
It is a little more fork, because for each new database I have to
create new schema and execute the repo_* scripts but I like it.

Regarding the ashmon utility it is a great tool. I was always feeling
that Oracle does not pays much attention to wait interface and I could
not understand why. That has changed in version 10g but was true in
version 9 and below. As I started my career on version 8 and 9 I was
miising that a lot. Around 5 or 6 years ago I found very easy
interface that visualise the wait statistics - rrdora -
http://home.zonnet.nl/r.vermeer/rrdora/ . It was already someting and
at least I could see the waits that were happening on the database,
with nice and simple pictures, without having to launch statspack, run
traces or looking at v$ views.

Although the rrdora could show me the waits it did not show which
actual statements were causing them. I could only guess, then run the
statspac, then guess again. Ashmon with s-ash is amazing tool that
actually lets me to see what SQLs are causing the waits. I believe it
should always be that way. It is great that we have the that
funcionality in 10g, but it is even better that I can have similar
capabilities on 9i. Thanks again.

What I find strange in ashmon is:
1) It often happens that I do not see the sqltext of the selected
querry. I have noticed that most of the time it helps to run
sash_pkg.get_all on the target host. As it is run every hour it is
sometimes too seldom to see the sqltext of recent sql.

2) It often happens that SQLs visible have 0 values in SQL_ID, SQL
type and plan_hash. It looks like those are the background processes
like LGWR, etc. but it is only a guess.

3) As I run the ashmon with s-ash on 9i databases I believe that the
explain plan functionality is not available for me (I do not see it
anyway)

4) Clicking on waits window gives me:
working …
wait_analysis: event :log file parallel write:
wait_analysis: id log file parallel write
wait_analysis:toto log file
or shows a SQL:
–ashmon
select
sum(cnt) over ( partition by ash.sql_id order by sql_id ) tcnt, (…)

what is the point of that?

5) Clicking on user_name window gives me:
working … session id 10,929
but nothing happens.

I suspect that some of the problems are related to 9i dtabase and the
fact, that I actually connect with ashmon to repository database and
not actual database, but just wonted to note the spoted issues.

Thanks again. It is great to know what is going on in my database
right now, and few hours before without the effort to go through
statspack guessing and tracing burden.

You are the man. If you ever visit Poland (where I am living) consider
yourself invited for a party.

Greetings
Remigiusz Boguszewicz

http://www.perfvision.com/ftp/ashmon_linux32.tar.gz

still needs a small change to soft link tile0.7.8 to /usr/share/tcl :
cd /usr/share/tcl8.4
ln -s /opt/ashmon/lib/tile0.7.8 .
lrwxrwxrwx 1 root root 24 2008-10-14 12:27 tile0.7.8 -> /usr/share/tcl/tile0.7.8

Thanks to Simon Coter from Oracle Italy who got it working during my 2 day class this past week in Milan.

Thread 2.4
The thread source code is here:http://www.perfvision.com/ftp/tcl/thread24.zip
You need to compile it for your machine.
If you already have the thread 2.4 library and ASHMON isn’t finding it then you can force load it with
load full_path_and_name_of_thread_file
Like on Solaris I could do
load $env(MON_SHARED_LIB)/libthread2.4.so

Tile 7.8
Should be available here: http://rpm.pbone.net/index.php3/stat/4/idpl/4616762/com/tcl-tile-0.7.8-alt1.i586.rpm.html

thank you for the answer.

I tried at first with standard debian provided packages.

apt-get install tk8.4 tcl8.4 blt tklib

then with the ones provided at your site. At both cases I end up in
the following error:

invalid command name “thread::names”
while executing
“thread::names ”
(procedure “connect” line 35)
invoked from within
“connect ”
invoked from within
“.m.f.o.connect.f.b.ok invoke”
(”uplevel” body line 1)
invoked from within
“uplevel #0 [list $w invoke]”
(procedure “tkButtonUp” line 7)
invoked from within
“tkButtonUp .m.f.o.connect.f.b.ok
”
(command bound to event)

To be honest I have no idea what could I try next.

I would greatly appreciate any suggestions.

Greetings
Remigiusz Boguszewicz