2008/12/24 xufeng <xuf...@yuanjie.net>: > > >> -----Original Message----- >> From: baron.schwa...@gmail.com [mailto:baron.schwa...@gmail.com] On Behalf >> Of Baron Schwartz >> Sent: 2008年12月24日 22:06 >> To: Jake Maul >> Cc: mysql@lists.mysql.com >> Subject: Re: On fighting with master-slave replication lag >> >> On Wed, Dec 24, 2008 at 2:31 AM, Jake Maul <jakem...@gmail.com> wrote: >> > Slightly more complicated (and also probably more accurate- the time >> > reported by show slave status is known to be unreliable in some cases) >> > would be a script that inserts a row into a table, then check the >> > slave over and over till it arrives. Or even better, insert 2 >> > values... a timestamp that *you* provide (in a shell script, something >> > like $(date) would work) and a timestamp generated by MySQL.... >> > assuming the times are syncronized on the master, slave, and the box >> > you're inserting from, when the insert hits the slave it'll generate >> > it's own timestamp, which you can then subtract *your* timestamp from. >> > >> > There's also a tool in maatkit which does replication tracking, >> > although I've not yet used it. Judging by the other tools in that >> > package though, it's probably pretty decent :). >> >> It is mk-heartbeat, and it does pretty much what you described, >> although it's been tweaked to be slightly more complex to suit various >> real-world scenarios. >> > I have read some stuff on http://www.maatkit.org/doc/mk-heartbeat.html and > am interested in this tool. I guess in reality the mk-heartbeat tool checks > the output of show master status on the master with focus on the File and > Position fileds.
If you really read that link, it puzzles me how you could come to that conclusion about the tool. It does no such thing and I think http://www.maatkit.org/doc/mk-heartbeat.html#DESCRIPTION describes that pretty clearly. Let me know if the documentation needs to be clarified.