next »

[nutballs]

lol

so my Master Master array just went in the shitter.

weeeeeeeee funz!

Apparently, round robining is not so smart.
stuff got so out of sync due to index collisions that I think it just told me to fuck off.
So now I am back to 1 server.
I was getting pounded to hell by google. It must have decided that my spamcontent was like ambrosia and was tearing up my sites.
The replication could not keep up, and eventually there was a collision which shut off the slave for about 8 hours until i noticed this morning. At that point there were so many replication errors because both databases were being written to, that it just shit itself when I tried to resync it.
SET GLOBAL SQL_SLAVE_SKIP_COUNTER = 1000000;  seems a bit.... bad to have to do.

[perkiset]

LOL your system is a trip meng.

Any progress on the clustering thoughts? I'm still curious about dedicating 4-6 of my V20s to a mini-cloud and seeing how that would work... I wonder if that would do you as well?

[nutballs]

i would love to.

I am figuring I would run the authority on my webserver and then the nodes would be the two database servers. The authority would only be used by the nodes, but never by the applications.

I have not been able to find a tutorial I can understand though... lol

[perkiset]

Same. Still too complicated for the amount of bandwidth I have to give it ATM. But I covet that. I think. 

[nutballs]

i covet lots.
especially if the machines dont need to be equal...

Im sure its not tough, but I just don't have time either.

So for now, I am back to 1 server as master, 1 plain slave which of course can become master in a crash.
I guess this is good anyway, since I can then use the slave for dumps, without bringing down live.
If i ever get around to clustering, I will probably drop in a shitty little box as a slave for that purpose.