Active IQ Unified Manager Discussions

How to simulate and measure the perfomance impact of a node-ctl failure within an HA in AIQUM

JavierBarea

Hello people

i was wondering after a customer sent a doubt to me, if AIQUM is able to simulate and measure the impact of CPU/IOPs that can handle one of the HApair in case of failure of the partner.

 

I would like something exactly or pretty similar to this that exstsed in OCPM (Perfomance Mgr) long time ago where you can assess this behavior.

 

see this to compare & respond to me

https://www.youtube.com/watch?v=42o4BOiajg0&list=PLdXI3bZJEw7mp3jFSg0IJgo3oeDQRY-Gl&index=21

 

Thanks in advance

 

 

1 ACCEPTED SOLUTION

niels

Short answer: Yes it does.

 

This is a screenshot from current AIQUM 9.9 (also existed in earlier versions).

As you can see there is a tab called "Failover Planning" covering exactly this use case:

 

FailoverPlanning.png

 

 

 

 

 

 

 

 

 

 

 

 

 

 

To get there go to

Inventory

> Storage

> Nodes

Under "View" select "Performance - All Nodes" - click on the node of interest.

 

Kind regards, Niels

View solution in original post

4 REPLIES 4

niels

Short answer: Yes it does.

 

This is a screenshot from current AIQUM 9.9 (also existed in earlier versions).

As you can see there is a tab called "Failover Planning" covering exactly this use case:

 

FailoverPlanning.png

 

 

 

 

 

 

 

 

 

 

 

 

 

 

To get there go to

Inventory

> Storage

> Nodes

Under "View" select "Performance - All Nodes" - click on the node of interest.

 

Kind regards, Niels

JavierBarea

thx for the quick answer Niels,

perfect

BestRegards

paul_stejskal

One thing that you can consider too is there is a bit of headroom in the CPU failover with increased latencies up to about 70%/70%. It's pretty hard to keep it right at 50/50 (or some split totaling 100%) all the time, so you do have a bit extra to use. There are some background processes that run but given the combined CPU space onto a single node, if they run at all during failover, they get even less execution time.

 

That's a good ground rule that you can use in addition to the failover planning.

 

JavierBarea

Appreciated The feedback

Public