Being charged for dormant instances?

[GAEfan]

I just checked our instances. I have one instance that has been open
for nearly 4 hours, and served just one request. Another has been
open for nearly 7 hours, and served just 6 requests. We have others
that have served thousands of requests.

So far, that is 11 instance hours (88 cents and growing) to serve
those 7 requests. We are not going to pay 13 cents per request for
our hosting.

This is clearly a bug, as these instances should have been shut down
hours ago. Google, you need to fix this before you start charging,
else you are ripping off your customers.

[de Witte]

Same problem, our instances are most of time idle. I think the pricing is reasonable okay but the scheduler is causing these huge rates.

[Mark Bucciarelli]

i'm pretty sure this was covered in this forum already.

if i understand it correctly, the role of an idle instance

is (mainly) to spin up a new dynamic instance.  the more

idle instances, the more quickly ones that actually handle

requests will spin up.

m

--
You received this message because you are subscribed to the Google Groups "Google App Engine" group.
To post to this group, send email to google-a...@googlegroups.com.
To unsubscribe from this group, send email to google-appengi...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.

[Jason Collins]

We are seeing 600 Total Instances, but only 150 Active Instances (this
application is still set for Automatic Max Idle Instances).

I'm really, really hoping that the sample billing Instance-Hours
computation was made using the Total Instances and that scheduler
improvements combined with the Application Settings "sliders" will
make this a much more reasonable number.

j

[Jeff Deskins]

I had same issue with always-on feature.  It had 3 resident instances up with no traffic and a new request came in and it spun up another instance to serve the new requests while the resident instances were still idle.  My app only gets heavy traffic on certain days - so not sure if the scheduler thought it needed to get ready for next spike or what.

I have since changed the max idle instances from automatic to a number that is more realistic during those days.  That has helped to keep the number of extra instances down.

Jeff

[Ugorji Nwoke]

What? This doesn't even make sense. The role of the scheduler is to schedule a request or spin up a new instance. An instance does not exist to spin up a new instance. 

[Sandeep dey]

Same Problem Here, as you can see three of the resident instances don't have any load while
3 dynamic instances take all of the load. This is going to imply a double cost for us.

Seems like its a common problem and it might mess up a lot of bills. Will google kindly fix this up or 

atleast explain to me why there are 3 idle instance  and 3 active ones (I can understand one 1 idle instance)

Sandeep

[Gregory D'alesandre]

This is indeed an issue with showing the comparison bills while always-on is still running.  The Resident instances that you see are the ones that always-on keeps running.  Under the new model if you have a paid app, you will choose how many instances you want to keep idle at a minimum.  So, on your comparison bill it likely says you are paying for 3 more instances that you will in actuality, so it is somewhere in the neighborhood of $3.84 more per day than it will be in actuality (2 more instances running all day, I expect you'll still want to keep 1 running all the time).  We are working to fix this on the side by side bills.

Greg

--

You received this message because you are subscribed to the Google Groups "Google App Engine" group.

To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/pNA-QEBSxGcJ.

[Gregory D'alesandre]

Er, sorry, I meant $1.92 more per day than it should be.

Greg

[Nick Rudnik]

There seems to be some favoring of dynamic instances taking load over the always on which is also what I am seeing in my experience. I seem to always have the 3 resident instances taking no requests while additional dynamic ones are doing all the work. Could this be exaggerating the price estimates even further? I'm guessing that means my price estimate under the new model is probably 72 instance hours higher per day than it should be. 

[Gregory D'alesandre]

Yes indeed, you are correct.  The new scheduler is geared up for the new model and as such doesn't use the always on instances as much.  So it is probably 72 instance hours per day higher than it should be.

As long as we are talking about this, someone also reminded me of is that I haven't been stressing often enough that Max Idle Instances is your friend if you want to save money.  As Jon outlined in another post on the group, it does 2 things:

- Suggests to the scheduler the maximum number of idle instances that should be running

- Caps what we will charge you for in terms of idle instances.  This basically means even if your instance graph shows 40 idle instances, if you have it set to 1, we will only charge you for 1.  An useful formula to understand is: billable instances = min(active-instances + max-idle-instances,
total-instances)

Let me know if you have additional questions about this.

Greg

On Mon, Sep 5, 2011 at 8:09 PM, Nick Rudnik <nru...@gmail.com> wrote:

There seems to be some favoring of dynamic instances taking load over the always on which is also what I am seeing in my experience. I seem to always have the 3 resident instances taking no requests while additional dynamic ones are doing all the work. Could this be exaggerating the price estimates even further? I'm guessing that means my price estimate under the new model is probably 72 instance hours higher per day than it should be. 

--

You received this message because you are subscribed to the Google Groups "Google App Engine" group.

To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/ALXAlfAClZ8J.

[Jon McAlister]

Hi GAEfan,

Instances that have not received a request in the last 15 minutes are
completely ignored by the billing formula. They may be still running
[if hosted on fortuitous machines], and as such are technically
beneficial to you in that you avoid extra loading requests, but we eat
that cost. We only consider instances that have received a request in
the last 15 minutes for our billing formula.

> --
> You received this message because you are subscribed to the Google Groups "Google App Engine" group.