Chronos response becomes very slow when there are a lot of tasks

[857851000]

We use Chronos to schedule more than 1000+ tasks, and as the number of tasks increases, we find that Chronos has a very long response time when modifying, deleting, and adding tasks.By looking at the Chronos log, we can found a large number of same logs:

08:49:59.152 [pool-4-thread-1] INFO org.apache.mesos.chronos.scheduler.jobs.JobScheduler - Reloading jobs 08:49:59.590 [pool-4-thread-1] INFO org.apache.mesos.chronos.scheduler.jobs.JobScheduler - jobsToRun.size=1, jobsNotToRun.size=1001

In Chronos source code file JobScheduler, there is an infinite loop function mainLoop (),we tried to add sleep to reduce the cycle interval and found that the Chronos response time was greatly reduced. So can you consider adding a parameter to control the sleep time length according to the actual application needs,or give other better solutions.

In addition, when there are a large number of tasks Chronos Web page open time is very long or can not open.

Chronos version:3.0.1
Mesos version:1.1.0

[brndnmtthws]

What exactly did you change in the code to improve Chronos? What's the period of your jobs? Chronos should 'go to sleep' unless there's something to schedule.

[857851000]

By looking at the source code JobScheduler,we think condition.awaitNanos (nanos) should let Chronos wait for task scheduling at first,but it seems not work.So we force to add sleep at the beginning of the loop,the specific location is as follows:

Most the period of our jobs are 1 hour and 24 hours.

[brndnmtthws]

Can you run a test and print the values of conditionNotified.get() and nanos in that context? Also maybe print a log message every time notifyCondition() is called to see if that's the problem.

[brndnmtthws]

Ah, okay, I see a potential edge case. Let me throw up a patch, and hopefully it fixes it for you.

[brndnmtthws]

I'm not sure if that patch will work, so reopen this if it does not.