Oh, she scales

256 views
Skip to first unread message

Brandon Wirtz

unread,
Jan 7, 2013, 3:02:55 PM1/7/13
to google-a...@googlegroups.com

We are doing load testing.

 

image001.png

Mahron

unread,
Jan 7, 2013, 4:07:26 PM1/7/13
to Google App Engine
Oh, What a Beautiful Sight !

Kristopher Giesing

unread,
Jan 7, 2013, 4:54:50 PM1/7/13
to google-a...@googlegroups.com
16s average latency though...

Mahron

unread,
Jan 7, 2013, 6:30:30 PM1/7/13
to Google App Engine
What are the instances doing in term of read/writes and cpu stuff ?

Jeff Schnitzer

unread,
Jan 8, 2013, 9:30:58 AM1/8/13
to Google App Engine
On Mon, Jan 7, 2013 at 1:54 PM, Kristopher Giesing <kris.g...@gmail.com> wrote:
16s average latency though...

...and 294 instances at an average 0.175 qps.  Ouch.

Jeff

alex

unread,
Jan 8, 2013, 10:34:02 AM1/8/13
to google-a...@googlegroups.com
looks like it's a crawler so there's not much you can do when external
sites respond slowly, except for fanning out.
> --
> 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.

Brandon Wirtz

unread,
Jan 14, 2013, 1:09:49 PM1/14/13
to google-a...@googlegroups.com
Pretty much. And we check robots before, so that slows things down. But
this is not user facing, This is a background process that happens in
response to user actions.

Igor Kharin

unread,
Jan 14, 2013, 4:41:15 PM1/14/13
to google-a...@googlegroups.com
Not much? With crawler I'd rather "fan in" in MapReduce Framework's
MutationPool style. It's a waste of instance time to make whole thread
wait for just a single website.

Jason Collins

unread,
Jan 24, 2013, 12:47:50 PM1/24/13
to google-a...@googlegroups.com
We are doing a data migration between apps (moving our services layer (and data) to a different app than our user-facing app).
Screen Shot 2013-01-24 at 11.45.30 AM.png
Screen Shot 2013-01-24 at 11.46.11 AM.png

Jason Collins

unread,
Jan 24, 2013, 2:06:58 PM1/24/13
to google-a...@googlegroups.com
Probably worth mentioning that on the way up that slope, we still encountered a number of 10-s "we can't get an instance for you" errors - around 4% of the requests failed like this. Our instances were set to 3-Auto and our latency was set to Auto-1s; I was really hoping that the scheduler would keep up with these settings, but it appears not.

Once spun up however, it's rocking.

j
Reply all
Reply to author
Forward
0 new messages