0

Configuration:

Netapp (SAN) providing a 5GB FC LUN to VMware hosts (datastore).

(2) Server 2012R2 datacenter VMs configured with a 5GB shared vmdk using paravirtual SCSI controller with SCSI Bus sharing set to "Physical" for hosting data.

Problem:

Whenever I put a file on the shared disk on VM1, the data does not show up on VM2. I have to reboot VM2 in order for the data to show up.

I am requiring this configuration for a failover cluster for my subordinate Certificate Authority servers thus the requirement to have data show up immediately.

Alibi:

Please request for additional information that may help your understanding or to help provide a thorough solution.

CrashFive
  • 25
  • 1
  • 5
  • 1
    What filesystem are you using? It's likely that the filesystem is only designed to be mounted by one OS at a time. – EEAA Apr 18 '17 at 14:23
  • Did you set up a proper cluster filesystem? [Like this?](https://blogs.msdn.microsoft.com/clustering/2012/06/02/how-to-configure-a-clustered-storage-space-in-windows-server-2012/) If you're using a standard NTFS fs, I would expect to see exactly what you're experiencing. – EEAA Apr 18 '17 at 14:28
  • Talk to me about 'shared VMDK' here - what's doing the locking/clustering? – Chopper3 Apr 18 '17 at 14:36
  • @EEAA Q1: I am using standard NTFS. Q2: I was not aware of the necessity to configure a cluster filesystem for shared vmdk's; however, I'll look through your article and see if it helps. Thanks. – CrashFive Apr 18 '17 at 16:07
  • @Chopper3 Q3: I am not sure? Where would I look to find if something is being locked? Info: I followed this [link](http://searchvmware.techtarget.com/tip/How-to-set-up-an-ESXi-shared-virtual-disk) to setup shared vmdk. – CrashFive Apr 18 '17 at 16:09
  • @EEAA per your link instructions, would I create 3 separate LUNs to ESXi to have 3 separate datastores then 1 vmdk for each datastore OR would I simply create 3 separates vmdk's off 1 datastore? Everything is coming off of 1 SATA aggregate, btw. – CrashFive Apr 18 '17 at 17:25
  • @EEAA - I have found [this](https://communities.vmware.com/thread/493840) to backup your statement. [Cluster Shared Volumes (CSV)](https://blogs.msdn.microsoft.com/clustering/2009/02/18/deploying-cluster-shared-volumes-csv-in-windows-server-2008-r2-failover-clustering/) apparently is the method to solve my issue because, as you stated, NTFS is not fond of dual OS writes. Please feel free to mark my question solved with the appropriate answer, if not, I can do it later, but would like for you to receive the credit. – CrashFive Apr 18 '17 at 17:37
  • 1
    "not fond" = "immediate catastrophic corruption" – Chopper3 Apr 20 '17 at 13:19
  • To amplify, you'll find this problem with any filesystem (on any OS) that wasn't designed for clustering. It's easiest for filesystem designers to assume that the OS instance is the only thing with access to the disk, so that's why there are specialty filesystems specifically for clusters. – mfinni Apr 20 '17 at 13:55

1 Answers1

0

It sounds like you want to cluster your filesystem between two hosts. You seem to be looking for a Cluster Shared Volume, but that is typically supported for newer SQL clusters or Hyper-V. I don't know if it would be supported for Certificate Authorities.

Typically, a CA is configured on both hosts to be identical. They share a disk in an Active/Passive state so one host has full control of the disk while the other is showing it offline. If the witness host loses connectivity to the active host, it will activate the passive host and the filesystem will be available to the secondary host.

If you're looking for an Active/Active approach, you might be able to do it with MFCS using Cluster Shared Volumes, but I know MS doesn't support the configuration so I would not recommend it for production.

Jarrod L. J. Gibson
  • 531
  • 3
  • 18