maria-developers team mailing list archive

["Philip Stoev" <pstoev@xxxxxxxxxxxx>]

Arjen,

What do you mean by "can be tricky and more often than not causes hassles", 
and in general, what is the semantics of a DDL KILL operation? For example, 
are you allowed to leave temporary files and if not, who is responsible for 
cleaning them up -- the engine or the server?

I am afraid that a timing-based method can only be used to test a KILL 
during the main phase of the DDL, that is, when the temporary table or the 
new index are being built. However, it would not be possible to reliably hit 
any of the other phases, such as when the old table is being replaced by the 
new table, without some explicit instrumentation of the code.

Philip Stoev

----- Original Message ----- 
From: "Arjen Lentz" <arjen@xxxxxxxxxxxxx>
To: "maria-developers" <maria-developers@xxxxxxxxxxxxxxxxxxx>
Cc: "Philip Stoev" <philip@xxxxxxxxx>
Sent: Wednesday, July 07, 2010 3:03 AM
Subject: How MySQL/MariaDB handles aborted DDL


> Hi Philip, all
>
> Philip's bug on recovery from a kill -9 during DML reminded me... from  my 
> experience, aborting a DDL query (ALTER TABLE - things like adding  an 
> index usually) b.
> By abort in this case I mean killing the connection/thread that does  the 
> DDL.
>
> It would be great to somehow have a few test cases in the suite and it 
> needs to be tested with all storage engines.
> Provided the table it operates on is big enough, it should be possible  to 
> just zap it by timing.
> Thoughts?
>
> Cheers,
> Arjen.