Hi Reindl, Many thanks for your contribution. I did remember we discussed rsync backup method in the past and I was able successfully to do this on our staging 100Gig + MASTER<=>MASTER replication setup. The downtime was really minimum, minutes not hours. It used to fail on regular basis as we tried to fix the errors by skipping or any other ways, now it solid for 2 month with no major issued. Thanks, Egor
On 20 April 2012 13:09, Reindl Harald <h.rei...@thelounge.net> wrote: > > > Am 20.04.2012 13:44, schrieb nixofortune: > > Thanks, > > So the actions should be: > > 1. Stop slaves > > 2. On Slaves: SET GLOBAL binlog_format=MIXED > > 3. Start slaves. > > not "SET GLOBAL", put things you want in my.cnf > you want them also get active after restart :-) > > > What is "restart the replication from scratch with a binary ident copy > (rsync)" > > Is it to use rsync to rsync mysql data on masters and slaves? and than > run chnge the master to .... to start from > > yes, that's the way i init replications sicne years > > * stop slave > * "hot rsync from the running master > * stop master > * remove binlog files > * rsync again > * start master > * start slave > * "change master to...." > > this way has several benefits > > * if you are using de-duplication backups it may save space > * less possible errors because a binary ident start > > the downtime on the master can be reduced to a minimum if > you are doing the rsync on a local drive on the > master-server, starting the master and after that you > have all time you need to rsync the copy to the slave > while the master happily writes new changes in his binlog > and after starting the salve it will fetch the changes > > usually i have a shell-script for all actions > on the master machines which needs param "really" > or it would stop after the hot rsync and not stop > mysqld > > >
