Do you ever see large pending_ms values in your logs?

What does your code do between the capability checks and the memcache call?



Robert




On Tue, Mar 1, 2011 at 12:32, Cian Montgomery <[email protected]> wrote:
> Looking at my logs this morning for the nights activity I find a
> couple of troubling cases, like this:
>
> AppEngine Log: 2011-03-01 08:45:38.679 /reggie/dispatch 200 5879ms
> 3150cpu_ms 0kb Mozilla/4.0
> Appstats: 2011-03-01 16:45:37.204 POST /reggie/dispatch  real=1255ms
> cpu=0ms api=0ms overhead=8ms
>  @905ms capability_service.IsEnabled real=4ms api=0ms
>  @935ms capability_service.IsEnabled real=8ms api=0ms
>  @943ms capability_service.IsEnabled real=5ms api=0ms
>  @948ms capability_service.IsEnabled real=4ms api=0ms
>  @1187ms memcache.Get real=7ms api=0ms
>
> AppEngine Log: 2011-03-01 08:13:36.788 /reggie/dispatch 200 18016ms
> 3103cpu_ms 0kb Mozilla/4.0
> Appstats: 2011-03-01 16:13:34.758 200 POST /reggie/dispatch
> real=1816ms cpu=0ms api=0ms overhead=6m
>  @1455ms capability_service.IsEnabled real=4ms api=0ms
>  @1484ms capability_service.IsEnabled real=4ms api=0ms
>  @1488ms capability_service.IsEnabled real=3ms api=0ms
>  @1491ms capability_service.IsEnabled real=4ms api=0ms
>  @1727ms memcache.Get real=7ms api=0ms
>
> In both cases app stats under reports the latency reported by the
> Appengine logs and my code in the call does not get called for a long
> time after the apstats logging starts (905ms in one case and 1400ms )
> in the other.
>
> Are the machines hosting the apps overloaded? Are there other settings
> I can turn on to get more insights to these delays?
>
> Thanks,
>
> Cian.
>
>
>
> On Feb 28, 10:50 pm, Cian Montgomery <[email protected]>
> wrote:
>> Thanks for your response. I installed and have been inspecting my
>> application with appstats. I have noticed a comple of things, 1 the
>> latency times and log times of appstats do not match what is reported
>> in the log. consider the following 5 calls, Particularly the 3rd call
>> where the difference is over 400ms.  (log comparison is made more
>> difficult by time stamps not matching and the times of the log entries
>> being in different time zones.)
>>
>> 2011-02-28 22:10:55.336 /reggie/dispatch 200 45ms 116cpu_ms 0kb
>> Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_6_6; en-US) AppleWebKit/
>> 534.13 (KHTML, like Gecko) Chrome/9.0.597.102 Safari/534.13,gzip(gfe)
>> 2011-02-28 22:07:27.537 /reggie/dispatch 200 64ms 154cpu_ms
>> 60api_cpu_ms 0kb Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US)
>> AppleWebKit/534.16 (KHTML, like Gecko) Chrome/10.0.648.82 Safari/
>> 534.16,gzip(gfe)
>> 2011-02-28 22:07:08.333 /reggie/dispatch 200 473ms 177cpu_ms
>> 60api_cpu_ms 0kb Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US)
>> AppleWebKit/534.16 (KHTML, like Gecko) Chrome/10.0.648.82 Safari/
>> 534.16,gzip(gfe)
>> 2011-02-28 22:06:55.387 /reggie/dispatch 200 60ms 107cpu_ms
>> 60api_cpu_ms 0kb Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US)
>> AppleWebKit/534.16 (KHTML, like Gecko) Chrome/10.0.648.82 Safari/
>> 534.16,gzip(gfe)
>> 2011-02-28 22:06:54.179 /reggie/dispatch 200 52ms 130cpu_ms
>> 60api_cpu_ms 0kb Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US)
>> AppleWebKit/534.16 (KHTML, like Gecko) Chrome/10.0.648.82 Safari/
>> 534.16,gzip(gfe)
>>
>>  (12)2011-03-01 06:10:55.308 "POST /reggie/dispatch" 200 real=20ms
>> cpu=0ms api=0ms overhead=0ms (6 RPCs)
>>  (13) 2011-03-01 06:07:27.487 "POST /reggie/dispatch" 200 real=40ms
>> cpu=0ms api=0ms overhead=0ms (8 RPCs)
>>  (14) 2011-03-01 06:07:08.264 "POST /reggie/dispatch" 200 real=55ms
>> cpu=0ms api=0ms overhead=0ms (8 RPCs)
>>  (15) 2011-03-01 06:06:55.337 "POST /reggie/dispatch" 200 real=43ms
>> cpu=0ms api=0ms overhead=0ms (8 RPCs)
>>  (16) 2011-03-01 06:06:54.140 "POST /reggie/dispatch" 200 real=32ms
>> cpu=0ms api=0ms overhead=0ms (8 RPCs)
>>
>> - checking for each capability enable costs 2-4ms best case. It would
>> be nice to be able to check all caps in 1 rpc.
>> - latency to the datastore is highly erratic,  on subsiquent user
>> requests for the same data from the data store show the following
>> resutls:  real=7ms api=85ms,  real=305ms api=85ms. Is is normal to 40x
>> variance in executing the same query to the datastore?
>>
>> Is there a way to put my own events in the appstats? I am using a
>> dispatch pattern in the server, so it is difficult to devine what
>> which handler I am looking at in the stats?
>>
>> Thanks,
>>
>> Cian
>>
>> On Feb 28, 12:42 am, Robert Kluin <[email protected]> wrote:
>>
>>
>>
>>
>>
>>
>>
>> > Hi Cian,
>> >    Have you used Appstats to investigate?  Perhaps there is some type
>> > of cleanup process, maybe by a session library, happening?
>>
>> > Robert
>>
>> > On Fri, Feb 25, 2011 at 17:00, Cian Montgomery
>>
>> > <[email protected]> wrote:
>> > > My app (appid: mr-october, always on) is experiencing lots of problems
>> > > with latency problems over the last couple of days.
>>
>> > > For a bunch of the calls to the same action dispatch handler over the
>> > > last there hours.
>> > > 2011-02-25 12:36:09.581 /reggie/dispatch 200 4829ms 4763cpu_ms
>> > > 236api_cpu_ms
>> > > 2011-02-25 12:35:31.683 /reggie/dispatch 200 159ms 405cpu_ms
>> > > 125api_cpu_ms
>> > > 2011-02-25 12:35:16.764 /reggie/dispatch 200 354ms 615cpu_ms
>> > > 125api_cpu_ms
>> > > 2011-02-25 12:20:37.208 /reggie/dispatch 200 232ms 452cpu_ms
>> > > 125api_cpu_ms
>> > > 2011-02-25 12:19:28.348 /reggie/dispatch 200 6380ms 4600cpu_ms
>> > > 190api_cpu_ms
>> > > 2011-02-25 10:47:19.700 /reggie/dispatch 200 171ms 452cpu_ms
>> > > 125api_cpu_ms
>> > > 2011-02-25 10:47:15.865 /reggie/dispatch 200 141ms 405cpu_ms
>> > > 125api_cpu_ms
>> > > 2011-02-25 10:47:15.731 /reggie/dispatch 200 385ms 475cpu_ms
>> > > 125api_cpu_ms
>> > > 2011-02-25 10:37:15.031 /reggie/dispatch 200 177ms 522cpu_ms
>> > > 125api_cpu_ms
>> > > 2011-02-25 10:36:48.890 /reggie/dispatch 200 3550ms 732cpu_ms
>> > > 125api_cpu_ms
>>
>> > > Which is honestly much better then the past couple of days where saw
>> > > some times over 30000ms.
>>
>> > > Or these where appengine did not even get to my log statement so I
>> > > have no idea what the request was:
>> > > 2011-02-23 09:36:35.922 /reggie/dispatch 500 141862ms 1395cpu_ms
>> > > 65api_cpu_ms
>> > > 2011-02-23 09:00:21.024 /reggie/dispatch 500 102858ms 793cpu_ms
>>
>> > > I can reproduce this behavor just by hitting the same query from the
>> > > client, reliably 1 out of about every 20 exections of this action is
>> > > over 5 seconds of latency.
>>
>> > > What can be done to address these latency issues?
>>
>> > > Thanks,
>>
>> > > Cian.
>>
>> > > --
>> > > You received this message because you are subscribed to the Google 
>> > > Groups "Google App Engine" group.
>> > > To post to this group, send email to [email protected].
>> > > To unsubscribe from this group, send email to 
>> > > [email protected].
>> > > For more options, visit this group 
>> > > athttp://groups.google.com/group/google-appengine?hl=en.
>
> --
> You received this message because you are subscribed to the Google Groups 
> "Google App Engine" group.
> To post to this group, send email to [email protected].
> To unsubscribe from this group, send email to 
> [email protected].
> For more options, visit this group at 
> http://groups.google.com/group/google-appengine?hl=en.
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"Google App Engine" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/google-appengine?hl=en.

Reply via email to