One issue that I ran into was the way the system monitors for file changes.
It does not respect the ignore list from app.yaml, it watches for each
instance of the front ends and/or back ends, and it restarts all servers on
any change.

Our system has at least two watchers checking ~10,000 files per second
each. Servers are bounced when we do something like 'git status', 'git
push', ...

This was using a ton of CPU and slowing down our dev servers.

I modified the "watcher" file to ignore some common directory names (.git,
.svn, ...) and log when it is bouncing the server. Attached is the modified
file, use at will. Would appreciate any enhancements sent back. To use this
file, copy it over the dev app server's file. You have to replace on each
SDK upgrade. On mac, this can be done with:
cp mtime_file_watcher.py
/Applications/GoogleAppEngineLauncher.app/Contents/Resources/GoogleAppEngine-default.bundle/Contents/Resources/google_appengine/google/appengine/tools/devappserver2

Jacob


On Fri, Aug 16, 2013 at 9:01 AM, PK <p...@gae123.com> wrote:

> Hi Mike,
>
> I have noticed that too. The CPU shows 25% even when the server is idle.
> It is not as bad as it was at the time they introduced the log server APIs
> but it is worse than what used to be.
>
> As another data point, I sporadically keep track of the time it takes to
> run a suite of end-to-end tests for my app and the time has increased 75%
> since 1.7.0. I have definitely not added so many new test cases, hardware
> is the same etc.:
>
>
> Date:   Wed Aug 7 15:40:54 2013 -0700
>     Linux ro2 2.6.32-33-server
>     Python 2.7.5
>     GAE release: "1.8.2"
>
>     real        *26m49.428s*
>     user        12m47.280s
>     sys 3m29.460s
>
> Date:   Thu May 9 13:37:31 2013 -0700
>     real        *23m34.431s*
>     user        10m12.420s
>     sys 2m18.190s
>     Linux ro2 2.6.32-33-server
>     Python 2.7.3
>     GAE release: "1.8.0"
>
> Date:   Wed Jul 25 21:29:57 2012 -0700
>         Python 2.7.2
>         GAE release: "1.7.0"
>         real    *16m8.724s*
>         user    8m21.050s
>         sys     1m45.360s
>
> PK
> http://www.gae123.com
>
> On August 16, 2013 at 8:08:47 AM, Mike Knapp (m...@shoesofprey.com) wrote:
>
> Has anyone else noticed the SDK consuming a lot of system resources
> recently? My computer is grinding to a halt, with the local dev server
> taking 60% of the CPU even when it's just sitting idle. It's painful to use.
>
> This seems to be a recent issue after upgrading. I'm on a Mac.
>
> Mike
>
>  --
> You received this message because you are subscribed to the Google Groups
> "Google App Engine" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to google-appengine+unsubscr...@googlegroups.com.
> To post to this group, send email to google-appengine@googlegroups.com.
> Visit this group at http://groups.google.com/group/google-appengine.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>  --
> You received this message because you are subscribed to the Google Groups
> "Google App Engine" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to google-appengine+unsubscr...@googlegroups.com.
> To post to this group, send email to google-appengine@googlegroups.com.
> Visit this group at http://groups.google.com/group/google-appengine.
> For more options, visit https://groups.google.com/groups/opt_out.
>



-- 
Jacob Taylor
CTO, Kahuna
(650) 302-1864
www.useKahuna.com <http://www.usekahuna.com/>
http://www.linkedin.com/in/thejacobtaylor<http://www.linkedin.com/in/adammarchick>

-- 
You received this message because you are subscribed to the Google Groups 
"Google App Engine" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to google-appengine+unsubscr...@googlegroups.com.
To post to this group, send email to google-appengine@googlegroups.com.
Visit this group at http://groups.google.com/group/google-appengine.
For more options, visit https://groups.google.com/groups/opt_out.

Attachment: mtime_file_watcher.py
Description: Binary data

Reply via email to