maria-developers team mailing list archive

[Paul McCullagh <paul.mccullagh@xxxxxxxxxxxxx>]

Hi Sergei,

Thanks for the great explanation.

On Dec 1, 2009, at 2:50 PM, Sergei Golubchik wrote:

> Hi, Paul!
>
> On Dec 01, Paul McCullagh wrote:
> > Thanks for the heads-up on that!
> >
> > The code in question is as follows:
> >
> > #ifndef WILL_BE_DELETED_LATER
> >  /*
> >    for now, only InnoDB supports 2pc. It means we can always safely
> >    rollback all pending transactions, without risking inconsistent  
> > data
> >  */
> >  DBUG_ASSERT(total_ha_2pc == (ulong) opt_bin_log+1); // only InnoDB  
> > and binlog
> >  tc_heuristic_recover= TC_HEURISTIC_RECOVER_ROLLBACK; // forcing  
> > ROLLBACK
> >  info.dry_run=FALSE;
> > #endif
> >
> > What would be the affect of removing these lines?
>
> I'd better start from "what is the effect of these lines" :)
>
> 2PC (with a binlog) works as follows:
>
>   prepare in all affected storage engines
>   binlog write
>   commit in all affected storage engines
>
> XA recovery (with a binlog) works as follows:
>
>   ask storage engines about the prepared but not committed  
> transactions
>   ask engines to commit every such a transaction that is in a binlog
>   and ask engines to rollback other prepared transactions
>
> Now, if a binlog doesn't indicate a crash (last binlog file was closed
> properly, showing a normal shutdown procedure) or, for example,  
> there're
> no binlog files at all, the server cannot perform the recovery - it
> doesn't know what transactions should be committed and what  
> transactions
> should be rolled back. A transaction may be prepared in one engine and
> already committed (or rolled back) in another. In this case the server
> requests the user to make the decision, the user has to request an
> explicit commit or rollback with --tc-heuristic-recover command-line
> switch.

So if I understand this correctly, tc-heuristic-recover handles a  
situation that should actually never occur.

> But if there's only one XA-capable storage engine, it's always safe to
> rollback all prepared transactions, there can be no other engine that
> has them committed. That's what the ifdef-ed code does - if InnoDB is
> the only XA-capable storage engine on recovery without binlog it  
> forces
> a rollback of all not committed transactions, preserving the pre-XA
> InnoDB behavior.

So, in fact, we could change this code:

#ifndef WILL_BE_DELETED_LATER
  /*
    for now, only InnoDB supports 2pc. It means we can always safely
    rollback all pending transactions, without risking inconsistent  
data
  */
  DBUG_ASSERT(total_ha_2pc == (ulong) opt_bin_log+1); // only InnoDB  
and binlog
  tc_heuristic_recover= TC_HEURISTIC_RECOVER_ROLLBACK; // forcing  
ROLLBACK
  info.dry_run=FALSE;
#endif

to:

if (total_ha_2pc == (ulong) opt_bin_log+1) {
  tc_heuristic_recover= TC_HEURISTIC_RECOVER_ROLLBACK; // forcing  
ROLLBACK
  info.dry_run=FALSE;
}


> > And what do you think we would need to test when we take out the  
> > lines?
>
> Start a transaction, do some table writes. Crash the server during the
> 2pc, for example:
>
>  SET @@debug="d,crash_commit_after_prepare";
>  COMMIT;
>
> delete all binlog files (or only the last one), try to start mysqld  
> - it
> should abort with an error.

OK, thanks. I will give this a try.

> The error message tells what to do - I hope
> you'll find it clear enough (but please complain, if not).

Will do!

Best regards,

Paul


> > On Dec 1, 2009, at 1:14 PM, Sergei Golubchik wrote:
> > > On Dec 01, Paul McCullagh wrote:
> > > > This patch changes MariaDB code in one case:
> > > >
> > > > File handler.cc, Line 1591
> > > >
> > > > DBUG_ASSERT(total_ha_2pc == (ulong) opt_bin_log+1); // only  
> > > > InnoDB and
> > > > binlog
> > > >
> > > > This line has been commented out, because the assertions fails now
> > > > that PBXT supports XA.
> > > >
> > > > Although this assertion implies that not everything will work, I  
> > > > have
> > > > not found any problem with the stability of MySQL due to the  
> > > > addition
> > > > of PBXT XA.
> > >
> > > Eh, no, not exactly. This assertion was to make sure that the  
> > > ifdef'ed
> > > code only runs when InnoDB is the only XA-capable engine.
> > >
> > > As it's no longer the case, you need to remove the whole ifdef  
> > > block,
> > > not only the assert.
> > >
> > > > PBXT XA can be disabled by setting pbxt_support_xa=0, it is  
> > > > enabled by
> > > > default.
> > > Regards / Mit vielen Grüßen,
> > > Sergei
> > >
> > > --
> > >  __  ___     ___ ____  __
> > > /  |/  /_ __/ __/ __ \/ /   Sergei Golubchik <serg@xxxxxxx>
> > > / /|_/ / // /\ \/ /_/ / /__  Principal Software Engineer/Server  
> > > Architect
> > > /_/  /_/\_, /___/\___\_\___/  Sun Microsystems GmbH, HRB München  
> > > 161028
> > >      <___/                  Sonnenallee 1, 85551 Kirchheim- 
> > > Heimstetten
> > > Geschäftsführer: Thomas Schroeder, Wolfgang Engels, Wolf Frenkel
> > > Vorsitzender des Aufsichtsrates: Martin Häring
> >
> >
> >
> > --
> > Paul McCullagh
> > PrimeBase Technologies
> > www.primebase.org
> > www.blobstreaming.org
> > pbxt.blogspot.com
> Regards / Mit vielen Grüßen,
> Sergei
>
> --
>   __  ___     ___ ____  __
>  /  |/  /_ __/ __/ __ \/ /   Sergei Golubchik <serg@xxxxxxx>
> / /|_/ / // /\ \/ /_/ / /__  Principal Software Engineer/Server  
> Architect
> /_/  /_/\_, /___/\___\_\___/  Sun Microsystems GmbH, HRB München  
> 161028
>       <___/                  Sonnenallee 1, 85551 Kirchheim- 
> Heimstetten
> Geschäftsführer: Thomas Schroeder, Wolfgang Engels, Wolf Frenkel
> Vorsitzender des Aufsichtsrates: Martin Häring



--
Paul McCullagh
PrimeBase Technologies
www.primebase.org
www.blobstreaming.org
pbxt.blogspot.com