After the topology is killed it takes up
to Config.TOPOLOGY_MESSAGE_TIMEOUT_SECS seconds before slots
become available again. During this time bolts keep process pending
What would be the best way to force kill topology in situations where
timeout is high (like 15 mins, or more)?
ps. Just wondering, how unusual is it to have such a long timeouts? Is
there a way to set timeout per component instead of per topology?
Grokbase › Groups › Google › storm-user › December 2012