-
Notifications
You must be signed in to change notification settings - Fork 1
faq
A: We ran both, the Master and the Client, successfully on the following operating systems:
- Ubuntu 16.04
- ArchLinux
- OSX
- Windows 10
Since the cluster is written in Scala / Java every operating system should be supported (even Android).
Q: What happens if I somehow disconnect a client hosting a virtual machine from the cluster in a very ungraceful way?
A: The master node will recognize the unreachability of the node and will stop all executors that currently perform tasks on the node. All affected tasks will be resetted. Don't worry, apart from a increased runtime there is nothing that you need to be afraid of 😉
Q: What happens if I somehow disconnect a client hosting a executor from the cluster in a very ungraceful way?
A: The master node will recognize the unreachability of the node and will reset all affected tasks.
A: Well it depends whether or not the device is still running - if so, than the tasks will still be (locally) executed (of course only if there is a virtual machine on the master node). If you somehow manage to restore the network connection than clients can again join the cluster.
A: Well then you are f*cked 💩
Q: I have so much computing power, but after a few minutes there are still only X VMs running - why is that? 😢
A: The virtual machines are being started / provisioned one after another. You just have to be patient, eventually all available resources will be used :)
Q: My task manager claims that there is still much RAM unused, but the cluster reports that it cannot start more virtual machines - why is that?
A: Usually this is because the RAM is not actually free, but used for buffering / caching. If any application requests memory this buffering / caching region will be freed, but until that it looks used. In order to free those caches you can (depending on your OS) use the following commands:
- OSX: (as
root)purge - *UNIX: (as
root)for x in {1..3}; do sh -c "echo $x > /proc/sys/vm/drop_caches"; done