plumber

still no water at home...
hopefully, one of my collegue (Manuel) organised a beer party in the bank of zurich. Well, 8 o'clock, nobody around, I assume the beers he left have to be drunk before they got warm.

Parallely, I am trying to clone oracle homes with grid. No chance!

I could well try tomorrow, but I think I am better here sitting and playing with this grid control.

OK... let's try once again. Where is this damned stupid "clone oracle home" button?

I must tell you that I found 10g grid control not intuitive at all. It is miserable how much menus you have to browse until you find what you are looking for! If you tried OEMGC, you know what I am talking about...

What should I do? Give up or have a last try (and a last beer)?

guess what, I took a beer.

Ok, I found a link!
http://zsi-aix621.prod.zkb.ch:7778/em
/console/ecm/cloneHome/cloneWiz_SourceHome

but there is not the 10g agent oracle home I am looking for here! I am going to get nervous...

I wished I knew Doug email address! My partner by rac was such a good companion!

Well... I cannot get it working. I apologize...

Just to make fun I want to tell my close friend John just asked me if I know the difference between hysterical reasons and historical reasons. Yes, I know what it means. I just like playing with words, but I know I am handicaped in german and english.... Whenever I try, you just think I am mistaking ;-))))

Anyway, it is almost nine o'clock now, I really should go home now...

linux 64bits 10gR2 is out

one fresh download today oracle database 10g download

I heard that grid control 10g release 2 may be available only in 2006. Quite annoying, since the repository cannot be on a 10gR2 db...

just waiting

no water, no fridge, no phone, no internet

I just came home. While rebuilding the bathrooms, they did cut the phone (nobody knows about that), the DSL, there is no water. I have water in the stairway, but I just got locked outside yesterday night and add to hurl (ring is also disabled) to get a neighbour opening the door. I have no fridge, but hopefully, the outside temperature is cold enough for the beer. I will survive.

I restarted working this morning. A few new downloads : oem 10.1.0.4 aix 5L and critical patch update october 2005. Still waiting for 9.2.0.7 aix 4.3.3

Torgon

I will have a rest in Torgon/Switzerland with my family next two weeks. In my flat in Zurich, they are going to completely renew the kitchen and the bathroom, and two weeks without water (to drink) and cool beer (to take shower) is not possible

restore to a new host : nid++

Great challenge today: restore to a new host from a closed noarchivelog backup on tape library.

In oracle 8i and before, the only way to rename a database was to recreate the controlfile. In 9i, I could change it with nid, in 10gR2, I should never have a reason again to recreate the controlfile, because even the MAXDATAFILES and MAXINSTANCES could be changed. Change controlfile is bad. Once a collegue forget one file, once he noticed it, only months later, he wondered how to recover it. Create a controlfile is way to much sensible thing too do. It is almost as bad as changing the dictionary!

Well. How to do than?

Ok, I have a database called LSC67 on prod, I want to recover it to LSC66 in devl. I have NO ACCESS to production, but I have access to tapes (one may wonder if this is a good security practice...)

Let's start.

First, if LSC66 already exists, shutdown abort. Remove datafiles. Remove controlfiles. Remove redo logs.

Now I restore the data/control/redo files from prod:/dbms/oracle/LSC67 to devl in /dbms/oracle/LSC66.

First I rename the files, some are called systemLSC67.dbf. I do not want that...

find do the trick

find /dbms/oracle/LSC66 -name "*LSC67*" |
nawk '{printf "mv "$1" "; gsub(src,target);print}' src=LSC67 target=LSC66 |
sh

I must just change the DB_NAME in my parameter file (which already exists at my site), to reflect the prod controlfile dbname

startup quiet force nomount
alter system set db_name='LSC66' scope=spfile;
startup quiet force mount

now I generate some statements for dynamically renaming the files

set ver off emb on pages 0 newp 0 lin 9999 trims on head off feedb off termout off
spool /tmp/rename_LSC67_to_LSC66.sql
select 'alter database rename file '''||name||''' to '''||replace(name,'LSC67','LSC66')||''';'
from (
select name from v$datafile
union all
select member from v$logfile
)
where name like '%LSC67%';
spool off
spool /tmp/drop_temp_LSC67_to_LSC66.sql
select 'alter database tempfile '''||tf.name||''' drop;'
from v$tempfile tf
where tf.name like '%LSC67%';
spool off
spool /tmp/create_temp_LSC67_to_LSC66.sql
select 'alter tablespace "'||ts.name||'" add tempfile '''||
replace(tf.name,'LSC67','LSC66')||
''' size 128M reuse autoextend on next 128M maxsize 2048M;'
from v$tablespace ts , v$tempfile tf
where tf.name like '%LSC67%' and tf.ts#=ts.ts#;
spool off

ok, now I am in mount, I do a rename datafile and drop tempfile. after I open database, and add tempfile. I am not taking care of the original size and autoextend clause of the original tempfiles, just 128M next 128M max 2G.

set echo on termout on feedb 6
@/tmp/rename_LSC67_to_LSC66.sql
@/tmp/drop_temp_LSC67_to_LSC66.sql
alter database open;
@/tmp/create_temp_LSC67_to_LSC66.sql

now I nid

shutdown immediate
startup quiet mount restrict

nid dbname=LSC66 target=/

and I change the db name and open resetlogs

startup quiet force nomount
alter system set db_name='LSC66' scope=spfile;
startup quiet force mount
alter database open resetlogs;

metalink

I just remembered the woman who asked Lawrence Ellison why you need at least three days to get a qualified support representative when you open an iTar.

Since a few days, I have been wondering why a non-dba user was not able to do sqlplus in 10gR2. Well, I saw the bug 4516865 on metalink for 9.2.0.7 and 10.2.0.1, but did not agree that it is a "desirable improvement", imho it is a major lost of service.

Just a few minutes (seconds?) after I finished writing my iTar, severity 2, I received a call from Oracle Support.

Apparently a group of customers asked to prevent any non-dba user from using sqlplus on a database server (just on the client). This kind of answer I do not like. But we had a long talk, he finally showed me a workaround for 10.2.0.1 : doing

chmod -R 755 $ORACLE_HOME

before root.sh.

Well, this should have made me hurl. But ok, I can do this.

He also explained me, that I should escalate this bug if I want development starting working on it. Also there is another bug, 4533592

My satisfaction with this tar is good (4/5). I have been contacted immediatly, informed about the bugs, and informed about the procedure to escalate this case when I will require too.

My overall satisfaction with Metalink is about 3/5. But sometimes, it is really 0/5. The worst I have was :

lsc: how do you remove headers in the middle of your result in sqlplus ? I have a bug with set emb on and set pages 9999.
metalink: This is a bug in sqlplus. Workaround is to use Oracle Reports.

I could just have kill her! The "good" solution is "set pages 0", ref: set pages 50000

One also like this :

lsc: When I try to do an executable in scheduler I got an ora-27371 (AIX)
metalink: switch to a new platfom
lsc: @*#!!
metalink: It will be fixed in 10.1.0.4
lsc: no it is not fixed in 10.1.0.4
metalink: it is an internal bug
...months later...
lsc: still not fix in 10.2.0.1
metalink: development is working on it

well, I have always known that AIX is not a strategic platform and many shell scripts function utterly wrong... still very frustrating one!

Larry just told that woman : just drop me an email. Should I write Larry each time I do receive an idiot answer?

Oktoberfest

Just come back from my first visite to Munich/Germany, where there is a very big party called Oktoberfest.

There is only one size of beer : one liter. After a few beers, I went to my friend Uli and his family.

It was a hudge party and and I enjoyed it very much...