Elastic job schedules jobs using jobConfig.cpu/mem while mesos master validates using jobConfig.cpu + appConfig.cpu causing job constantly failing; The same is true for memory and disk.

[lizhanhui]

Please answer these questions before submitting your issue. Thanks!

Which version of Elastic-Job do you using?

Latest release, aka, 2.1.2

Expected behavior

Task should be distributed to mesos agent with sufficient resources.

Actual behavior

Tasks is constantly assigned to the same mesos broker with insufficient resource(CPU, memory)

Steps to reproduce the behavior

1. Start a cluster with two or more agents
2. Launch a task taking up part of resource of the first mesos agent;
3. Launch a second task whose job cpu requirement is less than the first mesos agent remaining while app.cpuCount + job.cpuCount exceeds the first mesos agent's resources available.
4. The second launched job keeps failing.

Please provide the reproduce example codes (such as github link) if possible.

Example jobs suffice.

[terrymanu]

Fenzo cannot support custom executor natively, see: Netflix/Fenzo#110

We will try to fix it later, thank you.