And there are no relevant line in the logs of any of those four processes? (I don't care about watchdog since it doesn't connect to ActiveMQ.) What about at DEBUG?
What connection URI are your clients using? On Feb 1, 2016 12:01 AM, "Aaron Hu" <liouheng_...@qq.com> wrote: > Scenario: > I have five process: guard, watchdog, mdp, ems and security. Each process's > work is as follows: > Gurad: start an ActiveMQ server; monitor process watchdog, mdp, ems and > security, restart these process if they shutdown abnormally. > Watchdog: monitor process guard, restart it if it shutdown abnormally. > Mdp, Ems and Security: send alive message to process guard every 5 seconds. > > Question: > Step 1: start all these process. (Now process guard can receive alive > message sent by other processes.) > > Step 2: kill process guard. (Then process ems, mdp and security loss > connection with ActiveMQ, after a short time, process guard is restarted by > process watchdog. Now mdp, ems and security reconnect to ActiveMQ because > of > failover function.) > > Step 3: kill process mdp, ems and security. (Then after a short time, these > processes are restarted by process guard. Now process guard can receive > alive message sent by these new started processes.) > > Step 4: kill process guard the SECOND time. (This time when process guard > is > restarted by process watchdog, NOT ALL the processes mdp, ems and security > can detect this restart event, for example, process ems reconnect to > ActiveMQ and process Guard can receive alive message sent by process ems. > But the other two process CAN'T detect that process guard has restarted and > still send alive message to the former connection. ) > > HOW COULD THIS HAPPEN? > (ActiveMQ 5.10, JDK 1.6) > > > > > -- > View this message in context: > http://activemq.2283324.n4.nabble.com/Connection-lost-when-ActiveMQ-shutdown-and-restart-twice-tp4706711.html > Sent from the ActiveMQ - User mailing list archive at Nabble.com. >
