Comments for Laurent Schneider

Comment on On star transformation by Joe Gharibian

Joe Gharibian — Fri, 24 May 2013 02:42:39 +0000

Thank in advance for answering my question.

I am mostly interested in this comment “that 11g no longer requires bitmap indexes for star transformation. When I substitute b-tree indexes for your bitmap indexes..”
We are converting our PK indexes to use bitmap, knowing that is what Oracle suggests in DW. And also to take advantage of the start formation during execution. Those this work as well for no bitmap indexes in PKs? Doesn’t Oracle still convert the B-Tree indexes – the row ids to bitmap at the execution time?
Which one is more efficient? If Oracle uses start schemas for b-tree indexes, should we still convert all our PK indexes to bitmap?
Also you need to have the database parameter star_tranformation = TRUE if we want star transformations to work correct? I turned it on for our Data Warehouse globaley and some queries ran a lot longer, that forced me to turn it off.
Thanks,
Joe

Comment on use cron to schedule a job only once by Laurent Schneider

Laurent Schneider — Thu, 23 May 2013 15:55:47 +0000

good idea, or maybe

55 17 23 05 * /bin/runyourscript ; crontab -l | grep -v DELETEME-001 | crontab

so it will autodelete

Comment on Delete one billion row by Sudhir

Sudhir — Thu, 23 May 2013 02:37:52 +0000

test.snum has any trigger? I remember Tom Kyte mentioning something about “restarting” of internal activity of trigger when consistent get failing or some such thing …
Or may be just blocking lock? unindexed FK?

Thanks
/S/

Comment on use cron to schedule a job only once by Sudhir

Sudhir — Thu, 23 May 2013 02:34:35 +0000

How about

15 14 15 05 * /tmp/run-my-job; mv /tmp/run-my-job /tmp/run-my-job-hold 2>/dev/null

it wont find it next time
/S/

Comment on Delete one billion row by Laurent Schneider

Laurent Schneider — Tue, 21 May 2013 05:52:58 +0000

I fail to understand how you could delete more than one row with where rowid=:B3 Probably you use some cursor loop and you should simply have fresh stats and use delete from test.snum where yyyymmdd=:B1

Comment on Delete one billion row by Haris

Haris — Mon, 20 May 2013 09:12:22 +0000

HI Laurent

I have a delete statement with rowid, the primary key and date value(yyyymmdd) in the where clause but somehow the statement is running more than 12 hours. Number of records to be deleted is 107262 out of 218967984.

delete from test.snum where rowid=:B3 and snum_id=:B2 and yyyymmdd=:B1

Id | Operation
————————————
0 | DELETE STATEMENT
1 | DELETE
* 2 | TABLE ACCESS BY USER ROWID
————————————
indexes: pk_snum_id (snum_id)
idx1_prod_id (prod_id)
idx2_mnr (mat_num)

Any advice.

Regards
-Haris

Comment on New environment for OCM 11g by Laurent Schneider

Laurent Schneider — Sat, 18 May 2013 05:45:35 +0000

you can do most tasks in command line anyway. It was a very long time I last deployed 10g oem and I was pretty lucky and not well prepared on the 10g installation. But it is just a web tool and you are doing database OCM not middleware OCM

Comment on New environment for OCM 11g by Yannick

Yannick — Fri, 17 May 2013 13:07:26 +0000

Hi Laurent,

Thanks for notification ! Do you think it’s an issue to even use Cloud Control 12c, I clearly do not want to move backward on this…

Thanks, Yannick.

Comment on Oracle Password Repository by Laurent Schneider

Laurent Schneider — Tue, 14 May 2013 12:40:53 +0000

Hi,

Not sure what you are looking for. If your requirement is hardcoded password, why would you need a password tool?

The Oracle wallet is free, and you can use multiple wallets of course (just by changing TNS_ADMIN)… you can not have indeed multiple users as you do use /, which is a limitation obviously.

- what is the point in completely hiding passwords from DBA’s who can access anything anyway?
Valid question. One point is traceability. Audit. You could have a policy that each user/admin have its own logging, and only the “application server” is using the technical account. It does not protect you from the DBA. But it protects the DBA from being guilty by default, doesn’t it?

Comment on Oracle Password Repository by Jan-Marten Spit

Jan-Marten Spit — Tue, 14 May 2013 09:52:46 +0000

i wrote this a long time ago, and deployed it again this year (2013) to save a site from hardcoded passwords. my current requirements are:

- hardcoded passwords in scripts not running as oracle user, but on the same machine as the db instance.
- more than a dozen passwords, multiple database instances.

if there are better (and free/cheap) solutions i would like to know.

regarding the above article and comments (there is a README in the opr source which is intended to be read)

- the passwords are NOT stored encrypted, only obfuscated. the security is in the access to the file, not encryption of the stored password.
- the password file, if you follow the instruction in the README, is only readable for the user who created the repos, eg ‘oracle’. although you can grant access to other OS users.
- you can easily use a single file for more than one connection (which i never succeeded in doing with oracle wallet, but again, if someone knows..)
- what is the point in completely hiding passwords from DBA’s who can access anything anyway?
- i agree that in some situations (only a single user using passwords) , using shell/env is just as convenient