Very high request latency periods on App Engine

[Alex Epshteyn]

Last week I started noticing periods of very high request latency in
my app lasting for many hours and even days. These are not correlated
to the traffic patterns in my app, and so I assume it's an issue that
affects all of app engine. Looking at the logs, it seems that some
requests are spending way too much time in the pending queue, despite
my app's max pending latency set to "Automatic – 200ms" (min - max).

I filed an issue for this: http://code.google.com/p/googleappengine/issues/detail?id=6447

This is pretty bad. Could anyone on the team comment whether this is
an App Engine problem or something wrong with my app's configuration?

Thanks,
Alex

[Brandon Wirtz]

What is the max idle instances?

If a request takes 1000ms, and you get 8 at the same time, with 1 Idle
instance and a 5s spin up you would have time outs with these settings.

"Safe Rule"

If Max pending latency is set to less than the average request time:
Idle instances must always be >= Spin-up time / Average request time

If pending latency is set to less than average request time, you will ALWAYs
have to spin up another instance if you get more requests simultaneously
than you have instances (unless you are on a concurrent platform)

-----Original Message-----
From: google-a...@googlegroups.com
[mailto:google-a...@googlegroups.com] On Behalf Of Alex Epshteyn
Sent: Monday, December 05, 2011 4:32 PM
To: Google App Engine
Subject: [google-appengine] Very high request latency periods on App Engine

Last week I started noticing periods of very high request latency in my app
lasting for many hours and even days. These are not correlated to the
traffic patterns in my app, and so I assume it's an issue that affects all
of app engine. Looking at the logs, it seems that some requests are
spending way too much time in the pending queue, despite my app's max

pending latency set to "Automatic - 200ms" (min - max).

I filed an issue for this:
http://code.google.com/p/googleappengine/issues/detail?id=6447

This is pretty bad. Could anyone on the team comment whether this is an App
Engine problem or something wrong with my app's configuration?

Thanks,
Alex

--
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.

[bFlood]

seeing the same thing on one of my older M/S apps. Nothing has changed
in code but most of the request have huge latency with most in the
pending_ms queue. I started messing with the sliders to see if
anything worked but prior to that (when its been busted for several
days), the sliders were both on full automatic. 4-7 instances running
at any one time and minimal traffic