RE: Oracle's Updates Subscription Service

[Mohan, Ross]

Title: RE: Oracle's Updates Subscription Service

Walt,

 

Any thoughts on the partitioning schemes I mentioned?

 

Wouldn't this reduce the load on the daemon?

 

In addition, might the daemon be processor affinitied or nice'd to a better level?

 

 

I smell a solution, mate.

 

 

I am most curious to hear your thoughts......

 

- Ross

-----Original Message-----
From: Weaver, Walt [mailto:[EMAIL PROTECTED]]
Sent: Thursday, March 01, 2001 6:09 PM
To: Multiple recipients of list ORACLE-L
Subject: RE: Oracle's Updates Subscription Service

Thanks for the input, Ross.

 

As far as data corruption goes, it's neither semantic nor physical (well, there can be some semantic corruption in certain circumstances, but they're being worked on. We call them "bugs" here.)   :>)

 

Our developers seem to have inadvertantly gotten a grasp on the concept of an atomic transaction, and have coded many, many of them. As you know, MySQL does not do referential integrity, so all of our referential integrity is done at the application level. What appears to be happening is that during periods of heavy load the MySQL daemon occasionally gets confused in the midst of an "atomic" transaction and parts of the transaction are committed to the database while parts aren't. In fact, during periods of heavy load the MySQL daemon just gives up the ghost and dies. This is what we believe causes the data corruption.

 

BTW, there's no concept of a rollback in MySQL, either. That can be a problem when a transaction doesn't complete.   :>)

 

Could we really be seeing semantic or physical corruption and not realizing it? Well, maybe. But, we've moved our biggest, most active customers from MySQL to Oracle and their data corruption problems disappear.

 

MySQL definitely has its place in the world, and it had a place in ours back before we had customers beating on our databases. But, we're growing up now, and like all grown-up guys we're looking for bigger, more expensive toys. Oracle certainly fits that bill.   :>)

 

--Walt Weaver

  Bozeman, Montana, USA

 -----Original Message-----
From: Mohan, Ross [mailto:[EMAIL PROTECTED]]
Sent: Thursday, March 01, 2001 2:21 PM
To: Multiple recipients of list ORACLE-L
Subject: RE: Oracle's Updates Subscription Service

Walt,

 

Love that last line. But, i'll resist the temptation to bite the bait.

 

For you other stuff, i've embedded some newbie comments

below...stuff you've likely seen, heard, or thought about before.

 

thx

 

Ross

-----Original Message-----
From: Weaver, Walt [mailto:[EMAIL PROTECTED]]
Sent: Thursday, March 01, 2001 3:27 PM
To: Multiple recipients of list ORACLE-L
Subject: RE: Oracle's Updates Subscription Service

In addition, since MySQL has no concept of a transaction or rollback, data corruption is a constant problem.  

 

||  Huh?  Semantic corruption due to lack of developer transaction control is one thing. Actual, physical corruption

is a CONTROLLER or DISK problem. Or izzit something else? Your developers should be able to manage

transactions on the client, hell, all Oracle does is put x-action control in the db so the developers need to

think less ( "less", not "not at all") about it......  If you are talking about a hardware problem, you gotta fix

that...if it is semantic..i.e. code stepping on 'in flight' data, have a real sit down with your developers.....