in 1.0.3 version. When many request come to hecotor pool . and more than one node have problem(doesn't send reponse but node is alive).
1 . application did't have socket connection
and i try this commnad ==> netstat -na | grep | wc - l ==> result is 0
2 . Thread used hector pool will be blocked ... and if application or cassandra didn't recover status. many thread will be Time_wait status.
3 . and hector pool status : blocked : 400, active: 0 : idle: 0
this is thread dump. all most thread are ...
"[CASSANDRA_JOB_WORKER-2]thread-91" prio=10 tid=0x0000000041e2a000 nid=0x7e01 waiting on condition [0x00007f56f6c3b000]
java.lang.Thread.State: TIMED_WAITING (parking)
at sun.misc.Unsafe.park(Native Method)
so. i guess that my application lose connection socket. but it didn't recover connection in many reason.
so i think. how about make new routing policy. standard is Blocked/client count