Re: FAS8020 Network Cabling Requirements

SkyMover · ‎2016-10-24

Hello All,

I'm very new to the NetApp product. I've been handed an as yet unconfigured system - a single head FAS8020 with 4 DS4246 enclosures. I'm struggling somewhat to understand what physical network connections are required. I believe the system I have is referred to as a 2-node FAS8020 switchless configuration

It has been installed with the following:

1. Node mgmt connection (1 per node) on e0m: I have IP addresses for these

2.10Gb Ethernet on (2 each node) for data on eoc/e0d

3. Nothing on e0f/e0e.

4. Nothing on e0a/e0b

My questions are:

1) For a 2-node FAS8020 Switchless (which is what I believe I have) do I need direct node to node fibre on for the private cluster connect ports e0a/e0b? The FAS8020 quick install guide suggests I do but google searches have also indicated that for a single head FAS8020, you don't as the nodes will communicate via the backplane.

2. Do I need 4 connections for the cluster management on e0e/e0f? Do the cluster management connections need to be on a different network / vlan to the node management connections?

3. How many IP addresses do I need for the cluster management? Id it 1 or does it need i per e0e/e0f port?

I've read the various documents: FAS80xx Express Setup Guide/Network Management Guide several times but I'm still a bit unclear on what is the best way of providing the network connections. I've also searched this forum but didn't find anything that anmswered my questions.

Any advice/guidance much appreciated.

Chris

aborzenkov · ‎2016-10-24

1. 8020 only supports single chassis HA pair (with exception of MetroCluster configuration where you use extra FC-VI adapter), so it communicates over backplane. If you have MetroCluster I strongly suggest to not do it on your own if you are new to NetApp.

2, 3. Cluster expects and sets up by default 1 common cluster management IP and one management IP for each node. You can have as many ports connected to management network as you deem necessary and IP will failover between them as needed. Whether management and data networks are separate is entirely up to you and your requirements.

SkyMover · ‎2016-10-24

Thanks for the quick reply - very much appreciated.

No, we are definitley not doing a MetroCluster.

One further question - if we are not going to use e0a/e0b for the private cluster network can they be used for site to site mirroring or are they reserved exclusively for node/node comms? It is very unlikely that we would add a 2nd controller head to this system.

Thanks again

Chris

aborzenkov · ‎2016-10-24

There are no reserve ports. There are recommended ports for cluster interconnect that are offered by default during setup, but even these ports can be changed. And e0M port has limited functionality and usually is used as node/cluster management. So assuming you completed initial cluster setup any other port usage is completely up to you.

sichiba · ‎2017-01-10

Hello there,

I have an issue with the cabling of netapp fas8020. U can see enclosed how we did the cabling, for that we fellowed the recommendations on the oficial guide. Now that I can see it I have some alerts on bothy controllers which u can find below. Plz we need some assistance on that. If you could ever tell me if the cabling schema is correct or not I will be very grateful to you. Thank you in advance for your help

C2NETAPPSAPINO> system health subsystem show

Subsystem Health

----------------- ------------------

SAS-connect degraded

Environment ok

Memory ok

Service-Processor degraded

CIFS-NDO ok

Motherboard ok

IO ok

7 entries were displayed.

C2NETAPPSAPINO> system health alert show

Node: C2NETAPPSAPINO

Resource: Shelf ID 2

Severity: Minor

Probable Cause: Disk shelf 2 is connected to HBA ports 0b and 0a which

are both controlled by a single chip.

Possible Effect: Access to disk shelf 2 will be lost in a single chip

failure scenario.

Corrective Actions: 1. Halt controller C2NETAPPSAPINO and all controllers attached to disk shelf 2.

Connect one of the paths to disk shelf 2 either to another HBA or available port C or port D following the rules in the Universal SAS and ACP Cabling Guide.
Reboot the halted controllers.
Contact support personnel if the alert persists.

Node: C2NETAPPSAPINO

Resource: Shelf ID 4

Severity: Minor

Probable Cause: Disk shelf 4 is connected to HBA ports 0b and 0a which

are both controlled by a single chip.

Possible Effect: Access to disk shelf 4 will be lost in a single chip

failure scenario.

Corrective Actions: 1. Halt controller C2NETAPPSAPINO and all controllers attached to disk shelf 4.

Connect one of the paths to disk shelf 4 either to another HBA or available port C or port D following the rules in the Universal SAS and ACP Cabling Guide.
Reboot the halted controllers.
Contact support personnel if the alert persists.

Node: C2NETAPPSAPINO

Resource: Shelf ID 3

Severity: Minor

Probable Cause: Disk shelf 3 is connected to HBA ports 0b and 0a which

are both controlled by a single chip.

Possible Effect: Access to disk shelf 3 will be lost in a single chip

failure scenario.

Corrective Actions: 1. Halt controller C2NETAPPSAPINO and all controllers attached to disk shelf 3.

Connect one of the paths to disk shelf 3 either to another HBA or available port C or port D following the rules in the Universal SAS and ACP Cabling Guide.
Reboot the halted controllers.
Contact support personnel if the alert persists.

Node: C2NETAPPSAPINO

Resource: Shelf ID 1

Severity: Minor

Probable Cause: Disk shelf 1 is connected to HBA ports 0b and 0a which

are both controlled by a single chip.

Possible Effect: Access to disk shelf 1 will be lost in a single chip

failure scenario.

Corrective Actions: 1. Halt controller C2NETAPPSAPINO and all controllers attached to disk shelf 1.

Connect one of the paths to disk shelf 1 either to another HBA or available port C or port D following the rules in the Universal SAS and ACP Cabling Guide.
Reboot the halted controllers.
Contact support personnel if the alert persists.