Im attempting to create a desktop pool (using Horizon View 6) for the first time and hoping someone can shed some light on the problems Im having provisioning.
Im basically creating a floating pool of 25 linked clones on a dedicated node using a W7 image that about 32 GB.
Poweredge R730
dual 14 core E5-2695
256 GB RAM
1 local SATA for ESX
2 local raid 5 SAS datatores using 3 NLSAS 6Gbps drives each
1 local SSD SAS MIX use MLC 12Gbs for host cache and VM Swap
PERC H730 with latest firwmare (25.2.1.0037)
ESXi 5.5
The VMs have 2 core and 2 GB RAM each.
When I provision the pool, I see the replica get created and the VMs getting created. As the VMs are getting created, the vsphere events tab start showing numerous connectivity issues such as "lost access to volume xxx due to connectivity issues" and then "Successfully restored access to volume xxx following connectivity issues." Once it restores access, I see more VMs get created, some get recofingured, some get powered on, etc... then it losses connection again and this process repeats numerous times. The connectivity errors happen on both the datastores and SSD volume.
I was orginally monitoring the host with vcenter but when this problem occurs, vcenter starts having problems accessing the host so I started monitoring the host directly. I was watching the performance graphs and after about half the machines are created, the graphs start sporadically missing datapoints for minutes at a time (regardless of which graph I am monitoring). After a few minutes datapoint start getting logged again. This happens sporadically until all 25 machines are completely provisioned.
I have tried creating the pool with the replica and clones assigned to the 2 datastores which seems to take about 70 minutes before all 25 machines are completely provisioned. I then tried with the replica on the SSD and clones on the datastores which took about an hour. Ive also tried switching changing the image from thin to thick provision but that did not appear to have much effect. Im not running any quickprep/sysprep. There are no resource reservations on the host. View is configured for the default of 8 concurrent operations.
Im wondering if its normal to take that long to provision 25 VMs. It concerns me that vpshere starts have connectivity problems. I understand that booting and configuring 25 VM is resource intensive but I thought vsphere was supposed to be able to handle that without the client becoming unresponsive. I tried it using a different identical R730 node and had the same problems. It does appear to recover but Im not sure its working the way it is suppose to. Perhaps this issue is better suited under the vsphere category unless this is normal for View.