Auto-scaling issue

[marcyn]

Hi CheckMates,

I have a lab box that contains MHO-140 and 4 appliances (2x 23500 and 2x 23800).
This box is used for a presentation for a new Customers.

Because on next month I have a conference where I will be presenting Maestro i decided that besides some general information about Maestro itself it would be good idea to just show audience how it really looks like and works live.
And it would be good idea to show the audience some cool features also ... which is for example the newest auto-scaling ability.

Because these 2 models (23500 and 23800) are not compatible with each other I have 2 SGs. In one SG1 I have 23500's and in SG2 23800's.
Each group has 1 appliance, and other two were set as "scale unit". Then in SGs configuration I've enabled auto-scaling which should work as follows:
- scale-up if Security Group load average is above 60% for 30sec
- scale-down if Security Group load average is below 20% for 30sec
So as you can see nothing extra ordinary.

Then I made a script that makes a huge load.
After execution of that script about 30secs later unassigned appliance was moved into SG - as expected.
Couple of minutes later it became ACTIVE.
So scale-up works flawlessly.

Then I stoped this script and load dropped to 0-1% - I could confirm this on "top" and on "asg perf".
However appliance remained in this SG ... even after 30 minutes, and even after reboot it is still in SG...
So it looks like scale-down doesn't work...

Because @Lari_Luoma is the alfa and omega in Maestro I asked him about this strange behavior and he confirmed that he tested auto-scaling and everything ws working fine (scale-up and scale-down).
He sugested TAC - which is of course resonable.

Unfortunately after I described this to TAC I received information that this case is not for TAC... a little bit strange info but we can leave it...
Maybe someone else from CheckMates faced this issue ?
Maybe I'm doing something wrong ?

I thought that it could be something broken in R81.20 Take10 ... so I decided to test the same but on 2nd SG where my appliances are pure R81.20 (without any Take) - unfortunately it looks the same.

--
Best
m.

[PhoneBoy]

Curious how this is not a TAC case when it sounds very much like it is.

[marcyn]

I couldn't agree more 🙂

If we are talking about a feature that looks like not working well ... it's definitely case for TAC but I received:

(...)

This is beyond TAC scope, I would request you to reach out to SE or PS team.

(...)

WTF ? 😉

Anyway I will not give up so easily 🙂

[the_rock]

@marcyn I probably got that EXACT response at least 10 times before...This is beyond TAC scope, I would request you to reach out to SE or PS team. Must be a script or something lol

[Daniel_Szydelko]

It's a good point to start SR escalation 🙂

BR

Daniel.

[the_rock]

I will say, in my own experience, when you would escalate the case in the past, you would get response in 20 mins tops...now, you escalate it, you are lucky if someone replies in 4 hours. Anyway, thats story for another post, but let us know @marcyn once you find a solution, please share it here mate.

Cheers.

[Daniel_Szydelko]

Unfortunately, true

[marcyn]

Right now I wrote my "thoughts" for TAC's answer ... and we will see how it will go 🙂

[the_rock]

Dont give up @marcyn 🙂

[marcyn]

Never !

It's not my first "battle" with TAC 🙂

BTW
Of course when I will finally resolve this scale-down issue I will let you know guys what was wrong.

[the_rock]

Ah, so you are an "frequent flyer" as they say, like me LOL

Alright, keep us posted.

Cheers,

Andy

[marcyn · Accepted solution]

Hi guys,

So it looks like that after my "complains" TAC figured out that it is case for them and gues what ? .... I already have the answer.

It works as expected - by design.
And in my opinion after you will see TAC's answer it is really great that it works like that ... and not as I expected in the first place.

Here is the answer (and solution):
"By design it should not be removed. It should stay in security group, and if other Security Group will exceed its threshold, it will automatically move free scale unit (floating appliance) to another security group."

And if we will think about it ...
Let's assume that after 30sec threshold for scale-up was fulfilled and Maestro added scale unit to Security Group, and then after next 10sec load dropped below threshold for scale-down...
So it is good designed in my opinion that this scale unit remains in Security Group until needed by another Security Group.

This has only one "flaw" - there will be more time reguired for this Security Group to receive this scale unit (first it should be removed from previous SG, then added to the new one) - but anyway I think that it's still better then "flip-flop" 🙂

So ... case closed - now you and me, we all know how it works.

--
Best
m.

[the_rock]

Good job!