Has anyone got FreeNAS working on VMware ESXi ?

[Keljian]

Smb, ya lost me there, as in CIFS? VMware doesnt support that as a stg protocol, there must be a mixup?

Yes CIFS(well I have been told that officially it is now called SMB, but whatever), no I am not sharing with ESXi, VMs are on ssds.

 

[Keljian]

Hardware here:

Constellation - My new home/home office server build

I have 8 drives now rather than 7 but that gives you an idea.

For reference FreeNAS is using 2 Vcores, 12 gig of ram

 

[whitey]

Yes CIFS(well I have been told that officially it is now called SMB, but whatever), no I am not sharing with ESXi, VMs are on ssds.

GOTCHA, so you have those VM's on SSD's and they are local datastores to ESXi? If so you should really see the value in getting those ssd's in FreeNAS as well and running on NFS/iSCSI shared/clustered datastore. If you just have one host maybe not as big a deal but fun to experiment with.

2vcpu/8GB of memory here, should probably bump my FreeNAS AIO's to 12GB memory...wouldn't hurt.

 

[Keljian]

I don't want esxi to have to rely on the FreeNAS VM, otherwise things like PFsense(and net access) or mythbuntu(thus free to air TV) could go down when/if FreeNAS goes down. That would be a very bad thing.

Part of the reason for putting things in VMs in the first place was to keep the functions separate. I used to do it all (file serving, services, etc) in a Linux bare metal install, now I have it all separated out. Even have a tiny Ubuntu server VM now which just operates as a MySQL low transaction database despite having lots of VMs I could put it in if I wanted to.

 

[whitey]

Fair enough, I am used to LONG NAS uptimes, this one is shorter due to co-lo maint but I've had them up for 1000+ days no issues.

[root@nas-lvco-pr01] ~# uptime
5:39PM up 277 days, 1:21, 1 user, load averages: 0.44, 0.44, 0.48

If and outage happens lmore frequently than the time I do major maint/upgrades I have to seriously reconsider my stg strategy.

 

[CreoleLakerFan]

My FreeNAS VM is pretty stable. I've had my AIO for about six months now, the only time any of it has ever gone down is when I powered down gracefully due to a power outage that my UPS couldn't outlast.

A nice benefit I discovered of putting a VMFS SSD datastore on FreeNAS/ZFS is automagical reduction of empty space in the vmdk's if you have compression set on the volume. Thick provision on ESXi + compression on ZFS nicely trims it down to what you are using only. I guess thin provisioning has the same effect, but somehow it gives me the warm and fuzzies.

 

[Keljian]

Fair enough, I am used to LONG NAS uptimes, this one is shorter due to co-lo maint but I've had them up for 1000+ days no issues.

[root@nas-lvco-pr01] ~# uptime
5:39PM up 277 days, 1:21, 1 user, load averages: 0.44, 0.44, 0.48

If and outage happens lmore frequently than the time I do major maint/upgrades I have to seriously reconsider my stg strategy.

Well, for example, if I decide to go to FreeNAS 10, when I do the upgrade, I don't want to have to take everything else offline.

Recently I upgraded a Ubuntu 14.04 VM to 16.04, and it failed catastrophically. I had backups, but I didn't want it to take everything else down.

I had recent issues with mythbuntu also, same deal.

I guess what I am saying is, the majority of times I need to bring a VM down are due to my own fault when trying to implement features or upgrades as opposed to an inherent problem with the platforms involved.

I am serving the home though, so I have a higher tolerance for problems than a production environment.

 

[Keljian]

Thick provision on ESXi + compression on ZFS nicely trims it down to what you are using only. I guess thin provisioning has the same effect, but somehow it gives me the warm and fuzzies.

Thin provisioning (I find) doesn't seem to reclaim space once used due to fragmentation. That said, this problem could be a lack of knowledge on my part

 

[CreoleLakerFan]

Thin provisioning (I find) doesn't seem to reclaim space once used due to fragmentation. That said, this problem could be a lack of knowledge on my part

No, I think you're right ... Which means you have one more good reason to put your datastores on zfs instead of attaching themm to ESXi.

 

[CreoleLakerFan]

Well, for example, if I decide to go to FreeNAS 10, when I do the upgrade, I don't want to have to take everything else offline.

Recently I upgraded a Ubuntu 14.04 VM to 16.04, and it failed catastrophically. I had backups, but I didn't want it to take everything else down.

I had recent issues with mythbuntu also, same deal.

I guess what I am saying is, the majority of times I need to bring a VM down are due to my own fault when trying to implement features or upgrades as opposed to an inherent problem with the platforms involved.

I am serving the home though, so I have a higher tolerance for problems than a production environment.

That is a relevant point. I have a pair of 80GB S3500's as directly attached ESXi datastores. I have my pfSense and FreeNAS VMs stored there (I mirror the vmdk's via their respective OS). All of the other VMs are stored on a mirrored ZFS set of SM843T's.