Homelab setup

hste · Sat Jun 06, 2015 3:53 pm

I'm planning to setup a homelab with starwind VSAN
I have 2 servers:
Mainserver:
2xL5640 XEON
192GB RAM
16GB usb flashdrive for esx
256GB SATA disk for vcenter etc
Perc 6i raid controller with bbu
4x2TB raid 10 (4TB for datastore)
6x1Gb NIC

HA slave for virtual SAN:
2xE5645 XEON
192GB RAM
16GB USB flashdrive for esx
2x2TB RAID0 for HA (4TB for datastore)
8x1Gb NIC

I'm thinking on setting up mainserver with the VM with Starwind using pci passthrough so I could use the RAID controller software inside the OS for replacing fault disk etc.
Is this OK or should I make vmfs-datastore for the starwind VM on the RAID?

Should RAID 0 be ok for the HA slave?

Should I reserve 1 or 2 nic's for iscsi ?

Should I set 1 nic for sync direct attached server to server without a switch between?

hste

bubu · Mon Jun 08, 2015 1:42 pm

They have a wonderful tech paper https://www.starwindsoftware.com/Starwi ... ctices.pdf Why don't you use it?

According to mentioned above I think:

It's a great choice to use separated storage for StarWind VM.

RAID 0 is a good choice as well, since redundancy will be ensured by HA

The more NICs for sync you have, the faster your HA. The less points of failure you have, the more reliable your configuration

Mon Jun 15, 2015 12:32 pm

Please let us know if previous post was helpful

hste · Mon Jun 29, 2015 10:56 am

Yes I have seen that paper.

I now have started with a 1 node solution. As I don't have a UPS at moment and its just for a homelab. When I get a proper UPS I'll test the 2-node HA
What I have done so far is to install vsphere 6.0 on the mainserver on the 16GB usb flash and with vcenter vcsa on the local ssd storage

I configured the Perc6i raidcontroller with 4x2TB raid10 and in vsphere sat it to passthroug

I made a hyperv-server vm with 30GB and 32GB RAM and 6 cores and did the windows update. installeled vmware tools on it and changed network to vmxnet3 and used netsh to configure network.
Added a little extra disk with pvscsi and used diskpart to set it online and format to ntfs so that I could change c drive to pvsci.

I then removed the extra disk, and added the raid perc6i card.
I added the disk with diskpart and did a ntfs format. Assigned letter E to it.

In vsphere I added a vswitch for 2NIC made another subnet for it to use for iscsi,changed to jumboframes and configured failover.
I added a extra vmxnet3 card connected to the new switch and gave it an ipaddress on the new subnet and jumboframes.

I then downloaded and installed the starwind software.
I installed the starwind managment software on a pc and configured it to use the raid on E: added a LSFS disk with 500MB and 5GB and dedup cache, and a flatfile with 50GB
I added the target on the esx with the iscsi network ip. Scenned for the devices and found both the disk I had created on the SAN.
I sat the SAN'vm to automatic start after reboot.

I then installed som VM's on it and everything went well.

One problem I have had is when rebooting the system (taking down th san vm last) is that I have to remove the VM's on the SAN from inventory before I reboot, if not I can't access the volumes. I have tried to remove the VM's from the inventory and do a rescan, but something is locking so it cant connect to the volumes.

Performance is OK as far as I have tested.

hste

Wed Jul 01, 2015 1:49 pm

Could you please describe this in more detail

hste wrote: One problem I have had is when rebooting the system (taking down th san vm last) is that I have to remove the VM's on the SAN from inventory before I reboot, if not I can't access the volumes. I have tried to remove the VM's from the inventory and do a rescan, but something is locking so it cant connect to the volumes.

hste · Sat Jul 11, 2015 1:23 pm

Vitaliq (staff) wrote:Could you please describe this in more detail
hste wrote: One problem I have had is when rebooting the system (taking down th san vm last) is that I have to remove the VM's on the SAN from inventory before I reboot, if not I can't access the volumes. I have tried to remove the VM's from the inventory and do a rescan, but something is locking so it cant connect to the volumes.

I found the problem to be my networksetup and routing on the esx. When I fixed the network the problem was solved,

I still have to rescan after solarwinds has finished mounting the volumes.

Is there a way to trigger a script after the mounting is finished so that I can run a vcli script to do a rescan automtic when the mounting is finished?

hste

Sat Jul 11, 2015 4:59 pm

SolarWinds?

hste · Sun Jul 12, 2015 9:52 am

Sorry I often mix up the names of solarwinds and starwind. It was starwind of course.

I found a solution to get the disk to get up after a rebbot. I added a line to /etc/rc.local.d/local.sh on the esxi:

(until nc -z <ipadress of starwind iscsi-target> 3260; do sleep 15;done;sleep 60;until esxcli storage core adapter rescan --adapter <name of hba for iscsi like vmhba35>;do sleep 60;done)&

This line will do a rescan every minute of the iscsi hba till it complete without error after it check that the iscsi is up on port 3260 of the starwind.

darklight · Wed Jul 15, 2015 4:15 pm

Wow, thanks Hste.

Nice idea!

Mon Jul 27, 2015 12:17 pm

Thank you for the contribution hste