High Availability using OnCommand

[emanuel]

Hello

I am revisiting a topic i touched a few years ago.

I am curious to see if anyone has used any of the OnCommand suites in a High Availability configuration ( MSCS, Veritas Cluster, or in a VMWare vCenter ).  Looking at the Field Portal using a few search keys did not find much and searching the Communities has not brought any direct hits yet.

I am working with a client that is expressing interest in keeping their several OnCommand servers online and are currently hosted on bare metal boxes.

Does anyone have any TRs or good use cases I can read up on?

[adaikkap]

Hi Emanuel,

     The below TR should give you complete info on MSCS and VCS.

This TR gives a step by step details on how to configure High Availability for DFM on Windows using Microsoft Cluster Server (MSCS) and on Linux using Veritas Cluster Server (VCS).

TR 3767-High-Availability Support for DataFabric Manager Server

Regards

adai

[mike_burris]

If I read your post correctly you're more asking for use cases and examples of folks actually putting those configurations into production correct?  On the flip side, if I did by chance did read it wrong, I found the following docs on how to do MSCS and VCS:

MSCS:
https://library.netapp.com/ecmdocs/ECMP1113536/html/html/GUID-D728CCFC-A55A-4075-A55D-026E80419567.html

VCS:

https://library.netapp.com/ecmdocs/ECMP1113536/html/html/GUID-76210EB4-894E-48C3-B2F3-893B2593D9EF.html

I've not tried it with OC, but I know there were no issues running OM/PM 4.x in an MSCS setup - worked (and believe it still is) fine.

Mike

[emanuel]

Hello

Thanks for the link.  One big takeaway i am seeing is that iSCSI is not supported in a VCS / OnCommand environment?

[mike_burris]

I'm honestly not sure, but reading the docs they do specifically call out FC and I saw no reference to iSCSI.  The possibility is there that it may work, but I'd contact support to verify (perhaps the docs just havent been updated?).

Mike

[kryan]

Unless this has changed in recent documentation, iSCSI is not supported for the OnCommand Unified Manager (DFM) Linux VCS clusters.

https://kb.netapp.com/support/index?page=content&id=3011237

Regarding the LUN type:

• For any Windows configuration (stand alone or cluster), either iSCSI or  FC LUNs are acceptable.
• For Linux stand-alone servers, either iSCSI or FC LUNs are acceptable.
• For Linux VCS clusters, FC is the only choice (currently).

[kryan]

From the OnCommand Unified Manager Core Package 5.1 "Installation and Setup Guide" p122:

Configuration limitations for the OnCommand Core Package in VCS

You must be aware of the configuration limitations before you install the OnCommand Core Package in a VCS environment.

The following are configuration limitations in VCS:

• Only two-node clusters are supported.

• The OnCommand Core Package in VCS is not supported on VMware.

• For shared disks, the storage back-end must be an FC-based storage only.

• iSCSI storage is not supported for the OnCommand Core Package in VCS.

[loudymanschwab]

My company runs OM/DFM 4.0.2d12 on MSCS with FC SAN.  Haven't had too many problems with it, but upgrades and such can be a bit tricky (as the install package does not understand it's running on a cluster and that services are therefore managed by MSCS).

[emanuel]

Hello all

Thank you for the feedback ... really good stuff.

I am going to make presentation next week on a couple of possibilities with HA OnCommand; the client is using three installations with Linux so MSCS will not likely be one of them which leads me to one of two possibilities.

1.     A baremetal based Linux two node system using VCS with shared FCP based LUN storage on NetApp.

2.     if FCP is not in their diet, then I will make suggestion for a VM Ware host on a vCenter managed system that has the ability to failover a VM if a VM is lost.

This client will also be migrating 7-mode resources to c-mode clusters over time and I will plan on presenting an installation for OnCommand Unified Manager in C-Mode "mode"

I will read your replies in depth tomorrow.  If you got more, I will take it 🙂

Big Thanks!

[STORAGE_CIT]

I think your only choice is number 2 (vSphere HA) because you could use a Netapp FC-LUN multipathed OR a VCS-Ressource (if you need the possibilty to created snap´s of this LUN)  ...

In my last project, I designed and operated one of the largest Single-Instance DFM-Installations (more than 200 Filers/Instance), which is still "running" on a VCS-Cluster with Netapp-FC-LUNs.

Based on this experience i can only suggest: If you have the choice, don´t do it (don´t even think about it...)

One Example: Absolutely NONE of our cluster-takeover (unplanned and the planned one´s as well) worked properly.