Re: Request for enhancement: my.cnf knob to force slave to process se rver's own binlog entries
Guilhem Bichot <[email protected]>
| Newsgroups | gmane.comp.db.mysql.bugs |
|---|---|
| Message-ID | <1083172081.13278.2.camel@localhost> |
Hi, We had been in touch on this one year ago. On Thu, 2003-02-27 at 15:00, Guilhem Bichot wrote: > Hi, > > > How-To-Repeat: > > The old master hostA machine is then scheduled for repair or replacement > > > > - we analyze and or rebuild the host as appropriate > > - then it is time to make hostA into a slave of hostB > > - we copy the most recent backup from the backup cluser to hostA [this was > > an image from hostB originally] > > - we then start mysql on hostA > > - we then configure hostA to point to hostB and hostB's logfile position > > at the point of the recovery backup > > - we then start the slave on hostA > > > > The problem we have is that hostA winds up skipping all its original > > updates that are in hostB's binlog -- basically all the application updates > > from the time of the backup to the time of the failover. > > > > Now we understand this skip is necessary for bi-directional slave updates > > hostA->hostB and hostB->hostA at the same time. > > Indeed. > > > What we are proposing is a my.cnf knob that says a slave server will > > process all binlog events, regardless of the original source server ID. > > [e.g. the first test in slave.cc:exec_event() that detects that the log > > entry was from our serverid]. We understand this will create infinite > > loops on circular slave configurations, and this knob should obviously > > carry that caveat. > > It's a good but dangerous idea as you explained. I will put this in our TODO. > Note that this new feature will be neither in 3.23 nor in 4.0 as this two > versions are frozen (only bugs are fixed in these). Good news: I have implemented this feature today, it will be in MySQL 4.0.19 (which should be released in a few days). It is called the --replicate-same-server-id option, it is exactly the knob you asked for. Does MySQL still work fine for you? -- __ ___ ___ ____ __ / |/ /_ __/ __/ __ \/ / Mr. Guilhem Bichot <[email protected]> / /|_/ / // /\ \/ /_/ / /__ MySQL AB, Full-Time Software Developer /_/ /_/\_, /___/\___\_\___/ Bordeaux, France <___/ www.mysql.com -- MySQL Bugs Mailing List For list archives: http://lists.mysql.com/bugs To unsubscribe: http://lists.mysql.com/[email protected]