Instance issues

[Tomas]

From yesterday I observe that app engine spins new instance which serves requests for couple of seconds and then kill it immediately again. Can't see anything in logs, the spin up time also raised from usual 5-6sec to 18+ secs. That basically means that user faces 20secs timeout on each click. And as usually, GAE status all green. What's happening?

[johnP]

Our app too is having user-facing 500 errors (does not hit our code), and other nasty serving issues.

[Lawrence Mok]

are you using java / python / go? 

it seems to me every time when someone reported a problem, it luckily didn't happen to me. So I wonder if usually GAE got problem for some accounts only. Are you a paid user?

[Tomas]

Yeah sorry - I'm using pure java (servlets/jsps) on HR datastore, the user facing app is on http://www.librarist.com/ (which is spring mvc deployed on appfog talking with app engine backend). All day today I see a lot of app instance to be started and killed within few seconds, none of them have more than 200 requests processed - usually my dynamic instances serve ~50k requests in their live.

And the best part is that google is going to charge me extra for the CPU (I know it's few cents but what an irony) which they waste on spining/killing those instances while they just could run one dynamic forever which is usually able to server whatever my app needs.

[Lawrence Mok]

sorry to hear about that.... by the way I am using only "go" for a few months and there was only 1 major down time back into late 2012. Never had problem with my instance.. at least they run for a few minutes before killed.

perhaps the stability issue is depending on the language? 

By the way, I have configured the performance settings like this as my traffic is quite stable:

Idle Instances: ( Automatic – 1 )

Pending Latency: ( Automatic – Automatic )

Wonder how you set it?

[Kinesh Patel]

We have the same issue - instances are getting killed and causing user-facing 500's. On python 2.7 paid app.

[vlad]

I am using python 2.7, threadsafe=yes, all pref settings = default
Instances rarely live beyond a few minutes causing lots of loading requests. I have not yet turned billing on. I wonder if this is specific to FREE apps. I have a different paid app which seems to be ok.

[vlad]

Forgot to mention I have a task chain running every 90 sec. Many of the loading requests occur when running those tasks.