Exploding amount of running Instances and Costs without visible reason

[schwendinger]

I have a problem with an increase of running instances (and billed
instances) on a GAE Python Application. The application is online
since two years and it was used to consume an estimate of 10
instances at the time. Since friday 20th of january 2012, the
instances jumped up to a peak of 200 instances and does not go under
60 instances.

compare the instances on the following graph from the admin dashboard:
http://dl.dropbox.com/u/1028875/gae-instances.png

Therefore the costs of the app increased to about 4 times higher than
before (which is really expensive right now)

I did not made any code-chances in the last months and there is not
more traffic on the application than usual.

Any Ideas?

thanks,
Manfred

[Simon Knott]

Apart from the initial spike, the actual billed instances looks exactly the same - have you manually specified the number of idle instances you wish to have, or have you still got the slider on automatic?

Cheers,

Simon

[manfred schwendinger]

I used to have all settings on "automatic" but I recognized, that most of my other projects do also have an increase in costs since jan 20th without having more traffic or changes in the settings :(

the new costs per day are really high for the traffic we serve. some examples:

project 1 used to have costs of $20/day. on 20th it jumped up to $87 and is no on $40,-/day 

project 2 used to have costs of $3/day. on 20th of jan the costs jumped up to $9/day

project 3 used to have costs of $16/day. on 20th of jan the costs jumped up to $30,-/day

and I do have some more projects with the same price-changes.

I recognized that the only project which is not affected from the cost increase is the only project which uses High Replication Datastore. Are there any known issues with Non-High-Replication Datastore?

It seems like the non-high-replication Datastore got much slower, therefore the CPU time for frontend instances increased 50 to 100% and therefore the costs exploded. 

[Brandon Wirtz]

Moving to HR would be a good thing for a lot of reasons.  But It is hard to know what is causing your issue without knowing if you are Java or Pyhon Thread Safe or not. 

 

The Edge Cache changed a couple of times in how strict it was with headers, that may have made some difference. (10-60%)

 

The scheduler could have been tweaked.. That has happened a few times. And your settings of entirely automatic would make you more vulnerable to those changes.

 

Several of us on the list, Myself included do consulting to help with these issues.  If you are spending $600 a month on hosting you might find our services worthwhile. 

 

But I would definitely encourage you to move to HR as a first step.

--
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/-/0YWUmishgMsJ.
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.