Answer
inline.RegardsLongda------------------------------------------------------------------From:Renya
nath N <[email protected]>Send Time:2015年11月26日(星期四) 00:21To:dev
<[email protected]>Subject:Supervisor terminated without raising any
errorHi, Can you ease tell me why Supervisor terminated abruptly without
raising any error.[Longda] There are a lot of reasons for terminating
supervisor by exception. I have met some before.1. In some linux environment,
start supervisor without "nohup" and redirect it to backend, when logout the
system, supervisor is likely to receive "nohup" signal and quit.2. When os is
out of memory, os is likely to kill supervisor. You can check kernel log to
check it.3. When disk is out of use, supervisor is likely to be crash
immediately.4. When supervisor's memory is exceeding cgroup's setting, but
storm doesn't support cgroup right now.5. OS's ulimit's setting is wrong,
especially cpu time/memory. You can run "ulimit -a" to check this.There are
some other cases which will also lead to supervisor be killed. but you can get
some signal from supervisor.log or supervisor.gc.log.1. when zk isn't on
service, but you can get connect zk error in supervisor.log2. when supervisor's
memory is exceed jvm setting, but you can get some information from
supervisor.gc.logIt wasworking erfectly two days before. Also I cannot acces ui
in the browser even its still running... [Longda] If the ui can't be access,
please do the following check(1) check ui process running or not.(2) check
nimbus process is on service or not? you can run several command to check, such
as storm active xxxxx(3) check ZK service.
