Re: multi-instance sync woes

Subject: Re: multi-instance sync woes
From: Chuck Hill <email@hidden>
Date: Sat, 28 Feb 2015 01:07:44 +0000
Thread-topic: multi-instance sync woes

Hi,

Returning to this old message.

On 2014-11-25, 10:38 AM, "OC" wrote:

Chuck,

thanks a very big lot!

On 25. 11. 2014, at 19:11, Chuck Hill <email@hidden> wrote:

There are no simple solutions.  You need to be aware of where this can happen in your app and code for it.

A thorough refactoring is exactly what I would very much like to do :)

Alas, most of the current code is pretty old, written years ago with a firm intention to run single-instance (and even without  WOAllowsConcurrentRequestHandling), and before I get the time to re-write the guts and do it right, I must make the current thing at least _somewhat_ stable and useable.

When I get the "condition 379. Optimistic locking: multiple transaction conflict detected" I do essentially nothing -- I just wait a random couple of tenths of second, and retry.

I have seen that, but very rarely.  If all of the connections to the data base use "/isolation=read_committed/locking=optimistic” you should not get this.  I have caused this by using FrontBase Manager or some other tool to look at the database as it used a different isolation level and / or (don’t recall) locking

I might be wrong, but I believe it happens whenever two separate instances happen to update both before they commit, e.g., A: update, B: update (same row), A: commit, B: commit.

Depends on the application logic how probable this is, but I believe in principle it can always happen, and I believe if it does, in this scenario -- with the "TRANSACTION ISOLATION LEVEL read committed, LOCKING optimistic" setting in both sessions -- the latter commit consistently yields "Exception 379. Optimistic locking: multiple transaction conflict detected".

It happens to me occassionally; far as I can say, the isolation level is all right. Of course, I may be wrong, as always.

No, you are correct. That is a transient failure causing by a commit race exactly as you noticed. Thank you for making me go back and look at that. I have now added code to my frameworks to detect this (so far just for FrontBase as I have no examples from other database), and retry the save. I do that in a loop and give up after N attempts. The retried save either succeeds normally, or fails with a standard optimistic locking exception that my other code already handles. Nice!

Thanks!

Chuck

By the way, leads me to another question -- is there a way to send SQL and read the result string? I actually wanted to send "show transaction;" and put the result to my application log, to be sure the setting works on the production server (where I can't check through FBManager). I have found I should be able to use ERXEOAccessUtilities.evaluateSQLWithEntity(Named) to send the SQL, but I can't see how to obtain the results?

Thanks a lot again,

OC

 _______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list      (email@hidden)
Help/Unsubscribe/Update your Subscription:

This email sent to email@hidden

Prev by Date: Re: in another EC, when objects transferred through GIDs, owning relationship does not delete objects?!?
Next by Date: concurrency and consistency
Previous by thread: Re: first version of deployment config scripts for Ubuntu/AWS
Next by thread: concurrency and consistency
Index(es):
- Date
- Thread