Master/Slave Replication with Down Slaves

[Chuck Syperski]

Hello,

I am attempting to use the com.mysql.jdbc.ReplicationDriver on a web application and everything works great when my code/system is on the "happy path". My reads go to the replicas, my writes go to the master. My problem is that I might be misunderstanding how the driver is supposed to operate when a slave goes down. I am current using tomcat to manage the connection pool and here is my config....

<Resource auth="Container"
driverClassName="com.mysql.jdbc.ReplicationDriver"
defaultAutoCommit="false"
initialSize="3"
logAbandoned="false"
maxActive="200"
maxIdle="10"
maxWait="10000"
name="jdbc/mydb"
removeAbandoned="true"
testOnBorrow="true"
type="javax.sql.DataSource"
factory="org.apache.tomcat.jdbc.pool.DataSourceFactory"
username="username"
password="password"
url="jdbc:mysql:replication://localhost:3306,2.2.2.222:3306/mydb?autoReconnect=true&amp;allowSlaveDownConnections=true&amp;readFromMasterWhenNoSlaves=true&amp;connectTimeout=15000&amp;socketTimeout=15000"
validationQuery="/* ping */ SELECT 1"/>


My issue is that:

1. Tomcat won't start when the slave 2.2.2.222 is down. Is there a setting to allow for this?

2. If a node goes down at any point does the driver blacklist them, because it seems if a kill a node that my application hangs on queries that are going to the read only replica. Is there a way to have it ban failed nodes?

I guess in general does Master/Slave Replication gracefully handle failed slave nodes?

Thanks