Hi!
On Aug 10, 2008, at 8:27 AM, Jeremy Zawodny wrote:
Anyway, we talked a bit about the replication protocol and possibly
moving to something more flexible like protocol buffers[1] or
Thirft[2]. That'd be handy, since I've occasionally had a desire in
the past to whip up a Perl client that can speak MySQL replication.
I want the data coming out of replication to be completely open. The
server should be using the same library that any other client can make
use of.
Right now I am experimenting with buffers, and they seem to do the
trick. If I can get a working prototype this week of how we can use
them in the server then I will have hit my target.
But then today we had a network glitch at work[3] that caused a
corrupted log entry to get written to the slave's relay log. THAT
SHOULD NEVER HAPPEN. There should be a simple checksum (CRC, MD5,
whatever) on the log event that the slave can check before
committing it to disk and expecting the SQL thread to be able to
parse it. Fixing that by hand wastes my time, and I really don't
like things that waste my time.
Agreed. We will also add an ordering ID so that we can fetch/reply
records independently.
Cheers,
-Brian
--
_______________________________________________________
Brian "Krow" Aker, brian at tangent.org
Seattle, Washington
http://krow.net/ <-- Me
http://tangent.org/ <-- Software
_______________________________________________________
You can't grep a dead tree.
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
