Comments - Basics of MySQL failover (using replication) - Everything MySQL2024-03-29T11:49:22Zhttp://everythingmysql.ning.com/profiles/comment/feed?attachedTo=3993569%3ABlogPost%3A509&xn_auth=noNot sure I agree with manual…tag:everythingmysql.ning.com,2009-10-22:3993569:Comment:6082009-10-22T18:05:49.000ZChrishttp://everythingmysql.ning.com/profile/Chris
Not sure I agree with manual intervention here. Heartbeat, keepalive or any other automated mechanism would be optimal for a few reasons:<br />
<br />
1. Lets say you don't want to be woken up at some awful time of night or morning.<br />
2. Lets say you are the only DBA and you are in a place with no internet access<br />
3. Lets say you have 50, 100, 1000, 5000 servers... multiply that by the number of probable system failures and you have an administrative nightmare.<br />
<br />
Automated fail over is key to limit customer…
Not sure I agree with manual intervention here. Heartbeat, keepalive or any other automated mechanism would be optimal for a few reasons:<br />
<br />
1. Lets say you don't want to be woken up at some awful time of night or morning.<br />
2. Lets say you are the only DBA and you are in a place with no internet access<br />
3. Lets say you have 50, 100, 1000, 5000 servers... multiply that by the number of probable system failures and you have an administrative nightmare.<br />
<br />
Automated fail over is key to limit customer facing downtime plus heartbeat and keepalive are great tools with the right implementation logic. we've found that fully automa…tag:everythingmysql.ning.com,2009-10-21:3993569:Comment:6042009-10-21T20:09:34.000ZSheeri Cabralhttp://everythingmysql.ning.com/profile/SheeriCabral
we've found that fully automated solutions end up having their own problems, and that the best solution for most customers ends up involving manual intervention. Basically, have a master/master setup, where there are service names (separate A records that can be moved). Writes go to only one server (the "primary" server) and there's a standby (the "secondary" server).<br />
<br />
If the primary dies, or needs to be taken offline for maintenance, the secondary server can be promoted easily by moving the A…
we've found that fully automated solutions end up having their own problems, and that the best solution for most customers ends up involving manual intervention. Basically, have a master/master setup, where there are service names (separate A records that can be moved). Writes go to only one server (the "primary" server) and there's a standby (the "secondary" server).<br />
<br />
If the primary dies, or needs to be taken offline for maintenance, the secondary server can be promoted easily by moving the A records. No application issues, no heartbeat issues (it's finicky software!), and it only takes about a minute to change networking on both servers.