In-band ACP functionality

[SpindleNinja]

On IOM6s that have been upgraded to supported firmware and configured to use in-band, as well as IOM12s. How does the SAS channel actually issue the reset to the SAS chip / IOM that has stopped responding?

I see in this doc https://kb.netapp.com/app/answers/answer_view/a_id/1029778 it states: "The Alternate Control Path (ACP) interface currently uses Ethernet-based network connectivity to perform various recovery tasks, such as expander resets and shelf power cycles. This feature was referred to as Out-of-Band (OOB) ACP. With ONTAP 9.0 and later, ACP uses available SAS data links to perform the same tasks. This feature is referred to as In-Band ACP. The ACP mode that is configured applies to the entire ONTAP cluster."  

But with the SAS cable plugged in to the SAS chip, and the SAS chip is offline.. how does it reboot it should it lock?

[wronald]

Hi @SpindleNinja,

What is your problem exactly?

Can you send me some logs to check?

You might be facing an issue with the firmware of the Shelf or the ACP, and need an update.

In this case, follow that method:

Step 1: Connect the necessary Ethernet cabling.
 
Step 2: Switch to OOB ACP.

storage shelf acp configure -channel out-of-band -is-enabled true

Step 3: Download the latest ACPP firmware version 0209 or later (0210+ is recommended)

run local; storage firmware download acp

Verify the status of the download by running.

storage shelf acp module show -instance

Step 4: Switch to in-band ACP.

storage shelf acp configure -is-enabled true -channel in-band

Also, you can collect some SAS shelf logs executing this command on the nodeshell in diag mode:

node*> sasadmin expander_cli 1b.40.A "ddump"

Hope that helps.

Ron.

[SpindleNinja]

no issue, just wondering how it works at a deeper level. 

[GidonMarcus]

hi @SpindleNinja

The two IOM modules in the shelves can communicate with eachother.and to the backplane that control the power distribution within the shelf. so even whole unresponsive IOM can be powercycled

https://kb.netapp.com/app/answers/answer_view/a_id/1029350

Gidi

[SpindleNinja · Accepted solution]

@GidonMarcus  so If IOM A locks,  the system sees this due to it showing as unresponsive and sends a flag to IOM B on that shelf  to power cycle it? 

[GidonMarcus]

Hi

As much as i know there's very few scenarios any automatic recovery via ACP will be attempted (exemple for a bad one in BUG 868319). i don't believe it include issuing a full shelf powercycle. but it is available by command (and so as other reset and power_cycle commands) if you wish to issue them manually.

Gidi