Absolutely solid platform on vSphere 6.5

[Tom85768]

I've been running a FreeNAS 11.0U1 on a Dell R730 for over a year and a half now. We have the machine entirely virtualised on vSphere 6.5 an provide it with 10 x 1.46Tb SSD Virtual disks (as VMDKs) for a RAID 10 Volume, these are not passthrough. This is around 8Tb usable space. This is served up as NFS to a large vSphere environment consisting of hundreds of virtual machines. It also is the destination for our environments backups. Since deployment I have had little reason to look at or monitor the FreeNAS server which has been entirely rock solid. Whatever concerns other users may have previously had about running this platform as a virtual machine, without passing through the physical disks do not play out or eventuate in reality. Just wanted to thank the developers for this fantastic platform which in a 24/7 business environment never lets us down when running virtualised on vSphere.

 

[kdragon75]

We have the machine entirely virtualised

So if that host goes down ALL of your VM on the FreeNAS backed datastore go down...

provide it with 10 x 1.46Tb SSD Virtual disks (as VMDKs) for a RAID 10 Volume

Uhh... That makes NO sense.

This is served up as NFS to a large vSphere environment consisting of hundreds of virtual machines.

You should be running iSCSI at that scale...

It also is the destination for our environments backups.

Please tell me you not saving your backups on a VM on your primary storage!

I'm glad its working well for you but you have your self backed into a corner... Some serious house of cards stuff here...

 

[kdragon75]

I don't even know where to start with the technical issues...the fact that your running striped mirrors on VMDKs or the fact that.. Yep I can't get past the VMDK part. I mean, you not even using RDMs! You say you don't bother monitoring FreeNAS but it wouldn't know if there WAS an issue! Everything about the disk subsystem has be abstracted away! No SMART, no SCSI sense codes, nothing! It's the perfect storm and a perfect example of an environment that I would never touch without plans to rebuild all of it within 3 months. Even then I wouldn't want my name anywhere near it when stuff hits the fan.

 

[kdragon75]

Also, I'm sorry but "hundreds of virtual machines" is not a large environment. Large is hundreds of hosts each with a few hundred VMs spanning multiple vCenters with redundant PSCs. What you have is a giant single point of failure for that entire business. I am honestly concerned for them.

Please tell me your just trolling us. If you are you win the internet.

 

[kdragon75]

I see you ignored all of the advice presented to you in prior posts. You also failed to do basic research about why ZFS or at least the base OS should be presented with raw disks. To answer that please refer to my previously written post. I hope for your company sake that you get your VSAN up and running ASAP. You also need to understand how the LSI SAS 3108 (perc h730) may or may not obfuscate smart data, affect disk timeouts, etc... as it is NOT capable of true passthrough. In addition to any controller delayed timeouts, ESXi will have its own, then the VM. This all added up could be full minutes. More than long enough for all of you VMs to timeout on their IO and drop the disk (inside the VM) causing untold corruption in guest filesystems all before FreeNAS even knows that a disk was dropped!

 

[kdragon75]

You bought the wrong hardware, never validated your solution and hacked something together to CYA.