On Wed, 2006-03-08 at 20:15 +1100, Nick wrote:
I'm trying to mirror a live server onto a slave server of identical hardware. Theory being that when the master server blows up I can change the IP of the slave, restore the db's, restart a few processes and make everyone happy again.
What is the best way of doing this? Can you literally rsync / (with the exception of perhaps /proc and /tmp) and create an identical copy of the master? Is there a better way to achieve this?
I've looked at using cluster suite but I don't think the instant(ish) and automatic failover justifies the cost of the shared storage.
hi nick,
kk I'm doing exactly what you are trying to accomplish I think.
say the power went off completely.... nothing
k...comes back on... I run the exact same daemon and scripts in 2 mirror machines. I run the daemon in the first machine...it does the checking, realizes the other machine is not serving so it takes the roaming ip address and runs w/it. Then I run the daemon in the other machine...it realizes that the other box has the ball so it assumes slave responsibilities which are to just keep checking on the master...if the slave sees where the master is having problems, it takes the roaming ip and runs with it.
while all this is going on...the master keeps syncing up the slave oh, bout ever 15 seconds... Not perfect like a big $$$$ real time warez but it's close enuf for a small place like this!
I'd say when the slave does take the ball...the whole process takes bout 30-40 seconds which I don't like but that's life... the router has to be flushed one way or other in the deal...that's bout 20 seconds on this cheepie I got because I have to warm boot it...but this whole process w/all it's shortcomings on the time deals works flawlessly tho.
John Rose