Switch the catch to Throwable. Uncaught exceptions in a scheduled

threadpool executor will not report or log uncaught exceptions but will
kill the thread causing unexplained timeouts in CS.
This commit is contained in:
Hugo Trippaers 2014-07-16 16:22:31 +02:00
parent dbc7d80329
commit e78bb1d225
1 changed files with 8 additions and 5 deletions

View File

@ -24,6 +24,7 @@ import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicInteger;
import org.apache.log4j.Logger;
import org.apache.cloudstack.managed.context.ManagedContextRunnable;
import com.cloud.agent.api.Answer;
@ -298,9 +299,10 @@ public class DirectAgentAttache extends AgentAttache {
} else {
answer = new Answer(cmds[i], false, "Agent is disconnected");
}
} catch (Exception e) {
s_logger.warn(log(seq, "Exception Caught while executing command"), e);
answer = new Answer(cmds[i], false, e.toString());
} catch (Throwable t) {
// Catch Throwable as all exceptions will otherwise be eaten by the executor framework
s_logger.warn(log(seq, "Throwable caught while executing command"), t);
answer = new Answer(cmds[i], false, t.toString());
}
answers.add(answer);
if (!answer.getResult() && stopOnError) {
@ -317,8 +319,9 @@ public class DirectAgentAttache extends AgentAttache {
}
processAnswers(seq, resp);
} catch (Exception e) {
s_logger.warn(log(seq, "Exception caught "), e);
} catch (Throwable t) {
// This is pretty serious as processAnswers might not be called and the calling process is stuck waiting for the full timeout
s_logger.error(log(seq, "Throwable caught in runInContext, this will cause the management to become unpredictable"), t);
} finally {
_outstandingTaskCount.decrementAndGet();
scheduleFromQueue();