eth0 hardware hang

11 VMs on a Dual Core CPU. Isn't this a bit to much?

Ehh, yes is a bit too much (is a 4 thread core) , but does not eat to much core power and each VM is configured with 512MB of ram and one virtualized core.

The problem is not the core, is the network that is hanging and the kernel does not know what to do with this hanging, or something is wrong with virtualbox network driver.

I watched the LED's on the network, and when the network is dead , for three seconds all status LED's on the network card are off, and 7-8 seconds light up and show network activity, and continue to do this infinitelly, but the rest of the machine work well the HDD led show normal activity.

Tomorow I will receive the new hardware for this machine, a Rizen 7 1700 (8 core, 16 threads) and I will see the comportament of the machine.

I guess you have the wrong expectations here. Even if you don't see a full utilisation of all cores, the CPU has to queue up all operations of all VMs and cause delays which can cause "hickups" or lags. That effect you see in dropped network can be one effect of that. Each packet that has to be send or received accross that nic has to be processed by unraid to transfer it to the correct VM/Docker. In an situation, where the process chain/queue is already filled up to an certain point it can lead to a drop of the packet and an rerequest, depending on the protocol or worst case in an reset of that device. You will definitly see better performance with a newer higher core count platform, thats true, but even on a 8core CPU 11 VMs, all running at the same time is a lot of IO operations you're generating. Sharing cores between VMs and Dockers, running at the same time can and will affect the performance of all components and devices used by these VMs/Dockers. Seperate everything as much as you can to reduce side effects like stutter or lag

I kill three of the VM's, and now is working OK without any hang, I see that there is a limitation of maximum two virtual cores per processor thread, that I see it explicitly documented on Synology DSM that use the same VirtualBox and does not let you allocate more than double the number of threads of the processor, so the maximum viable allocated cores in my case is 8 virtual cores on a 2 core 4 thread procesoor.

About the wrong expectation is that I wondered if I can fool the machine to do this work until the new hardware will come from the store.

Ehh, my bad, I will change the status of this thread to "other" and "solved", and remain for others to see like documentation

Edited April 4, 2019 by MorgothCreator

Are you talking about Unraid or are you using Virtualbox or Synology DSM. These are 3 fundamental different software solutions. 

I talking about unRaid.

Virtual box is used in unRaid and Synology DSM, so the limitations are the same.
The only difference is that Synologi DSM does not let you allocate virtual cores more than double the threads of the processor.

You ever heared what nested virtualization is and what limitions there are? Virtualize something inside a already virtualized environment is NOT a good idea. You will face a lot of issues when it comes to performance. Why are you using a extra layer of virtualization, if you don't need to? To fool your machine it has more power than it actual has? Thats not how it works! 

All my VM's from above disertation are on a bare metal unRaid machine.

I only compared unRaid OS with Synolgy DSM OS that both use Virtual Box.

Who can think to virtualize inside another virtualized enviroment :))))

42 minutes ago, MorgothCreator said:

Virtual box is used in unRaid and Synology DSM, so the limitations are the same.

To sum things up, "Virtualbox" is NOT used in Unraid. Unraid uses Qemu/KVM which is a type1 hypervisor where VirtualBox is a type2 hypervisor. Thats 2 complete different things. 😉

Ahh, sorry, I am new to unRaid and I do not come to this part until now and I assumed that use VirtualBox :D, my bad.

But it seems that there are the same limitations

Small advice. Try to separate the VMs from each other. Try not to use the same cores on different VMs at the same time or you will end up with this situations you saw before.

 

Lets say you have 8 cores. Limit your total number of running VMs to max 7. Don't use core0, it is always used by Unraid itself to manage all the stuff in the background. And try not to split a core and it's hyperthread between 2 VMs. If you have a core+HT give it to the same VM. Sure it will work but you will drastically drop your performance and will end up with some weird errors. 

Ok, thank you for the advice