Scale IO 2.0 depoyment fails on nested esxi

Hi,

Please note such a configuration is unsupported and might lead to unexpected results.

Which exactly ScaleIO version are you trying to install? Can you paste the Virgo logs here?

Thank you,

Pawel

Hi Pawel ,

I am using EMC ScaleIO v2.0-12000.122 .  Isn't this possible on a nested ESXi environment ?

Thanks Hurse for the help .

I will try out on the physical servers , since its giving me so many issues .

Thanks Once again for the quick help .

As Pawel Said it is not supported, and you will definitely get unexpected results. I what you are experiencing is the type of issue I have run into on my nested lab.  it seems scaleio deploy script does not like the nested networking. It times out on the install failing certain steps. It could be different each time.

What you have to do is exit the install script I believe at the point you are at the SVM's should be deployed networking setup and passwords changed already. When you exit the install you need to shell into the svm's and manually install the cluster per the deploy guide. The install files would be in /root/install I believe.

Again the cluster will not react stable at all. It does not like the nested networking for I belive because of some possible latency or longer time it takes to respond.

Once deployed in the systems most likely you will see mdm switch overs and sds's going decoupled because of using this not tested nor supported config.

For a test lab setup we recommend a minimum of 3 physical esxi servers managed by vcenter, with at minimum one 1 gig network, however know that data traffic over a public 1 gig network is not the best idea for the data network. It would be suggested to have a separate data network specifically for between sds and between sds to sdc communication.

Shuhail,

AS the others have stated, it is not officially supported and definitely not recommended for a production environment.

That said, it is most likely a communication issue, so here are a few things that might help:

- Turn off the firewall on your nested hosts.

- Try enabling promiscuous mode on your vswitches.

- SSH in to the master MDM SVM and ping the IP of the slave mdm's. If it can ping, run the command yourself to see if there is a port blocked (6611 and 9011 on the MDMs)

Hope that helps

Ok sounds good.  You probably have it but suggest pulling the latest 2.0.1.X documentation set. Page 48 of the user guide does go over some of the networking designs.

https://support.emc.com/docu82880_ScaleIO_Software_2.0.1.x_Documentation_Set.zip?language=en_US

I haven't tried nested zen yet but I do have a virtualized Linux cluster that I can test commands and procedures and so on on it..  Seems I have ran into it only with nested esxi under workstation or esxi. Either way the supported deploys are physical and we can not promise proper results or expected behavior from a nested virtualized cluster. So don't rely on it to show stability and performance of the software.

We running scaleio on nested xenserver and linux without any problems. Not

try on vmware

So should i got ahead trying with the mentioned networking good practice or start off with the physical servers ?

Ricks response?

Mine was for physical..