Solved

Acropolis File Services Setup error and confusion

  • 21 October 2016
  • 20 replies
  • 2983 views

Userlevel 2
Badge +17
Hello Nutanix

I am currently in progress of deployment of Acropolis File Services (AFS)

Everything has been going well until the "Can't contact LDAP Server" error occured

I have given it a check by viewing some configuration on NVM and found out that although on Prism, they have been assigned with 2 IPs but actually running with only 1 IP when checked via console by ifconfig command (eth0 only)

Let me make it clearer via the "topology" below:

CVMs connecting to NVMs by Subnet A IPs
Clients, AD and DNS servers should be connected to NVMs by Subnet B IPs

Recently, those NVMs are running with only 1 Subnet A IP although on Prism, I saw they have already assign with both Subnets?!

My block is running with AOS 4.7.2. I have also given some seaches around and seen that both Internal and External networks could be the same as described in this video?!

https://www.youtube.com/watch?v=Q-z11wVhBxA

Thank you in advanced
icon

Best answer by hienle 26 October 2016, 08:47

View original

This topic has been closed for comments

20 replies

Userlevel 6
Badge +29
Can you submit a support ticket, so we can get on a webex and help you work through the issue?

Also, keep in mind that AFS is still tech preview in 4.7.2, so there's a few things that aren't perfect. AFS Will be GA with 5.0, which is coming out soon.
Userlevel 2
Badge +17
Hi Jon,

Thanks for your reply,

This AFS feature is recently being tested to prepare for upcoming evaluation so actually I need only some recommendation from you or supporters here to light up some solutions for solving the problems

Do you have any ideas beside open the WebEx session via creating the support case?
Userlevel 6
Badge +29
No worries, I understand.

I pinged a handful of folks from our AFS team to see if one of them can hop on here and help out.
Userlevel 6
Badge +29
Also, keep in mind that this is a community forum and has no formal SLA, so if you need more urgent help Support's your #1 route, and your #2 route should be your Nutanix account team
Badge +1
With AOS 4.7.2, it is not possible to deploy a fileserver with only a single VLAN. This capability will only be supported with our upcoming AOS 5.0 release.

However, you mention that you actually provided two different subnets, so it looks like the deployment process failed to bring up the second network, which caused the fileserver deployment to fail.

Do you know if the AD and DNS servers were also reahable from Subnet B? Our deployment process in AOS 4.7.2 expects that AD and DNS servers are reachable via both the Internal and External networks, and if not, then the process fails. We have eliminated this requirement in AOS 5.0, where there can be a clean separation.

I will be in touch with you directly to help you move forward.
Userlevel 6
Badge +29
I've got  on the case here, he will either respond here or email you directly.
Userlevel 4
Badge +19
Both subnets will need to be able to talk to the CVM. Is that the case?
Userlevel 2
Badge +17
Hi Jon,

Thanks a lot your heartily support, I will contact with  for further process of resoving this issue
Userlevel 2
Badge +17
Hi shubhankar,

Thank you for your consultation,

The AD and DNS servers are both deployed onto one VM. This VM is running on Subnet B and connecting successfully to other VMs that running on the same network

In my thought, I think because the NVMs did not bring up the second network interface that should be run on Subnet B so that was the reason why the errors come up




I have uploaded screenshots as on the Prism, the NVM is being shown up to be assigned with 2 IPs but actually when I connect to it for a check via console, it is running with one IP?!

The AD server is runing on 172.16.x.x subnet B
NVM is connected to CVM on 10.9.x.x subnet A
Userlevel 2
Badge +17
Hi dlink7,

Based on what I have known about AFS, there should be only one subnet for communication between NVMs and CVMs for file server cluster creation. Other subnet would be for connection from clients to file servers as well as between AD, DNS servers and NVMs.
Userlevel 4
Badge +19
Hi

If the default gateway for eht1 not reached then we wont show eth1 details. i would maybe check vlans and ping from fsvm and cvm.

Hope that leads you in the right direction
Userlevel 2
Badge +17
Hi dlink7

The connection between NVMs and CVMs is OK

Of course I cannot ping from NVMs to AD server as well as from AD server to NVMs

The default gateway for eth1 is not existed. Do you mean that the eth1 interface is actually is assigned and running but not shown up?

The AD server VM is running on Subnet B created on AHV server and connected to other same subnet VMs successfully
Userlevel 4
Badge +19
what network were you using for public interface? 10.9 or the same as AD?
Userlevel 2
Badge +17
The public interface is 10.9.x.x subnet (connected to Internet) as running for CVM and NVM communication
Userlevel 2
Badge +17
"Our deployment process in AOS 4.7.2 expects that AD and DNS servers are reachable via both the Internal and External networks"

Opps... after re-read this sentence from , I think the problem could be the connection between AD and Internal networks cuz the AD server cannot be reached by the 10.9.x.x subnet.

By the way, I have also review the AFS field in the Prism Console Guide but there is no point of mention about connection between Internal network and AD should be reachable or not. By the way, this would be cleared in the AOS 5.0, wouldn't it?

I will give it a test and come back whether it is OK or not.

Thank you alot for your support Nutanix guys
Userlevel 2
Badge +17
Hi all,

I have configured again so that both Internal network and External network can "talk" to each other and the problem has been resolved.

Thank you a lot for your advice  
Userlevel 2
Badge +17
Sorry for bringing this topic up again,

As I am testing the HA feature of File server cluster, whether the AFS cluster is able to provide the connection integrity to client as soon as AFS running on with 4 NVMs?

I have test HA with cluster running on 3 NVMs but failed and I have been having problem with my cluster now so that I could not test it on 4 NVMs.

Thank you in advanced
Userlevel 6
Badge +29
Are you asking if client connections will drop if you have one of the NVM's fail?

If so, no, they are not using SMB3.x, so they will have to reconnect just like traditional SMB shares (for now). Thats a future enhancement we're working on
Userlevel 2
Badge +17
Hi Jon,

You mean that currently, the FS cluster is just supporting for HA connection from SMB 2.1 and SMB 2.0...clients, not SMB 3.0, isn't it?

I am using Windows Server 2012R2 as well as Windows 10 clients to test so this should be the cause?

I am sorry if I could not get your meaning at all
Userlevel 6
Badge +29
For SMB3 to work correctly, both the client and the server need to support it. Of course 2012 r2 and windows 10 support SMB3.

AFS does not currently support SMB3 in the tech preview version that is available today