#NetApp has any tools for cabling with AFF ASA FAS series?

1 messages · Page 1 of 1 (latest)

topaz coral
#

NetApp has any tools for cabling (NVMe/SAS cables, FC cables, Ethernet cables) with AFF ASA FAS series?

reef reef
#

Generally the install guides cover off the install

#

click on install and setup A-Series systems

#

click through and you will get the PDF

#

eg

amber haven
#

And hwu has all supported cables and cards

topaz coral
#

example, each controller has 1 FC card with 4 ports. How I can connect to 2 SAN Switch for best practice?
As I think each controller go to both switches, port 1 and port 2 connected to SW1; port 3 and port 4 connected to SW2?

snow coyote
#

To be safe I usually assume that ports 1&2 are on the same ASIC. I then plug ports 1&4 into fc switch A and ports 2&3 into switch B. This helps with resiliency. If the card ASIC fails then you have reduced pathing to each switch instead of full loss to a single switch

topaz coral
#

do you have any document that said about which port will be same ASIC?

sand hatch
#

on the cards it's always neighboring ports. on the switch you'd have to ask Cisco, I guess it's 4 or 8 ports per ASIC

#

OTOH I have never seen a system where one ASIC on a card breaks, and the system doesn't panic/takeover, so I would argue it's more of a theoretical failure scenario

snow coyote
#

Depends on switch. Many of the newer Cisco switches have all ports on same ASIC
Nexus 9336, 93180 come to mind.

topaz coral
sand hatch
snow coyote
#

I found out middle of last year that the new Netapp SAS cards (x2071) have a single ASIC. When i have systems quoted i usually have a second sas card for resiliency installed

sand hatch
#

a single chip could still mean two separate ASICs just co-located on a single carrier. And even if you have two chips, the PCIe interface is usually again an SPOF. If you want to maximuze redundancy, always go with two cards. But then again, the PCI bus backplane might become your SPOF. It's just juggling numbers at this point, and you always have the fallback of HA that takes care of any failures

#

IMHO it's not worth trying to min-max the redundancy this way since HA works so well and takes care of everything that could go wrong hardware-wise