Someone is not sharing their unicorns. :(

On Thursday, December 27, 2012 11:32:01 AM UTC-8, Cesium wrote:
>
> FYI, my app is cruising along just fine with one instance that has been 
> alive for 2 days.
>
> I feel your pain, brother.
>
> David
>
> On Thursday, December 27, 2012 12:12:27 PM UTC-7, Carl Schroeder wrote:
>>
>> Also FYI, this whole issue could be resolved if you would stop sending 
>> user facing requests to cold uninitialized instances in java GAE. Handling 
>> user requests in ways that you know will cause 20+ second response times is 
>> pathological. 
>>
>> All new development for us on GAE is blocked until this issue can be 
>> resolved. 
>>
>> On Thursday, December 27, 2012 11:01:30 AM UTC-8, Carl Schroeder wrote:
>>>
>>> I am still seeing java instances decomissioned after sub minute quiet 
>>> periods.  Given that it takes 20-30 seconds to spin one java instance up, 
>>> you should probably leave them alive for a bit longer than a few seconds. 
>>> Otherwise, for low traffic profiles, page loads for GAE java can take up to 
>>> 30 seconds. God help us if the scheduler thinks I need 2 new instances spun 
>>> up in series rather than parallel.
>>>
>>> FYI, people don't wait around for a minute for pages to load. They use 
>>> other services.
>>>
>>> Once again, due to unannounced pathological behavior of the instancing 
>>> on GAE, we are wasting our time re-implementing our java infrastructure on 
>>> AWS. At least, I hope it is a waste of our time...
>>>
>>>
>>> On Tuesday, December 25, 2012 4:00:26 PM UTC-8, Francois Masurel wrote:
>>>>
>>>> Hi Saket,
>>>>
>>>> Thanx for investigating the problem.
>>>>
>>>> Things seems to have improved significantly these last few days.
>>>>
>>>> On our Neustar reports, for an average page loading, we can see that we 
>>>> went from 
>>>> 6.65s<https://monitor.wpm.neustar.biz/reports/graph_log/#/?monitorIds=8bc3b100-b372-11e1-a749-9848e1660ab3&timeRange=last_24_hours&functionType=loadtime&graphType=line_chart&view=loadtime>
>>>>  (12/16) 
>>>> to 8.94s (12/18) and back to 
>>>> 2.57s<https://monitor.wpm.neustar.biz/reports/graph_log/#/?monitorIds=8bc3b100-b372-11e1-a749-9848e1660ab3&timeRange=last_24_hours&functionType=loadtime&graphType=line_chart&view=loadtime>
>>>>  (12/23).
>>>>
>>>> I can confirm that we don't see as much instance warm ups in our logs 
>>>> as we were used before, we went from every 3-4 minutes to 20-25 minutes.
>>>>
>>>> I definitely think things have changed on Google side as we haven't 
>>>> changed anything on our side.
>>>>
>>>> But some strange things are still going on at the moment (12/26 0:52 
>>>> UTC+1) as a dynamic instance has been up for more than 4 hours but has 
>>>> served only one request (cf. screenshot below).
>>>>
>>>> Thanx again Saket for your help.
>>>>
>>>> François
>>>>
>>>>
>>>> <https://lh6.googleusercontent.com/-ARDAA2Yu-NY/UNo-Bgn9SpI/AAAAAAAAyN8/NNC4QLzi93g/s1600/Instances+-+VinoCities+-+Google+Chrome.jpg>
>>>>
>>>>
>>>>
>>>>
>>>> On Monday, December 24, 2012 11:54:48 PM UTC+1, Saket Kumar wrote:
>>>>>
>>>>> Hi Francois,
>>>>>
>>>>> Are you still facing this issue? I did a small test for your 
>>>>> application and didn't find anything too bad from scheduler's 
>>>>> perspective. 
>>>>> I'm trying to understand if it was a temporary glitch that was causing 
>>>>> the 
>>>>> issue or something is wrong with scheduler's algorithm. Or if scheduler 
>>>>> doesn't spin instances properly if QPS is low.
>>>>>
>>>>> Here is series of events- 
>>>>>
>>>>> a.) Single resident was serving for 1 day, 03:26:37 minutes, serving 
>>>>> 9600 request
>>>>> b.) After serving 96 requests with raised QPS, new instance was 
>>>>> created.
>>>>> c.) QPS lowered and the new instance was allowed to die.
>>>>> d.) Again the QPS was increased and new instance was created. Both 
>>>>> instances were handling requests at this point.
>>>>> e.) QPS lowered again, newly formed instance dies and older instance 
>>>>> starts serving 100% of the requests.
>>>>>
>>>>> -
>>>>> Saket
>>>>>
>>>>> On Saturday, 22 December 2012 12:18:00 UTC+1, Francois Masurel wrote:
>>>>>>
>>>>>> Hi Christina,
>>>>>>
>>>>>> For my personal case, my production app-id is "vncts1" with billing 
>>>>>> enabled.
>>>>>>
>>>>>> And the related issue is : 
>>>>>> https://code.google.com/p/googleappengine/issues/detail?id=7865
>>>>>>
>>>>>> As I said at the beginning of this thread, there are quite a few 
>>>>>> other threads related to this problem :
>>>>>>
>>>>>>
>>>>>> https://groups.google.com/forum/#!msg/google-appengine/cJcOrOE4JDc/uKFBbylFOxIJ
>>>>>>
>>>>>> https://groups.google.com/d/msg/google-appengine/8KiXRP-oU5I/Jf5kzcvyIiwJ
>>>>>>
>>>>>> https://groups.google.com/d/msg/google-appengine/NR_oV4znvks/KtkSmVQD2IgJ
>>>>>>
>>>>>> https://groups.google.com/d/msg/google-appengine/_znmrJwFMuM/V2epYOQ19L4J
>>>>>>
>>>>>> https://groups.google.com/d/topic/google-appengine/hg3BH8WOGWA/discussion
>>>>>>
>>>>>> https://groups.google.com/d/topic/google-appengine/UN5oCWkPsaA/discussion
>>>>>>
>>>>>> https://groups.google.com/d/topic/google-appengine/RneC9qDbeRE/discussion
>>>>>>
>>>>>> https://groups.google.com/d/topic/google-appengine/sLZcZoK-A3Y/discussion
>>>>>>
>>>>>> Thanx for you help.
>>>>>>
>>>>>> François
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Saturday, December 22, 2012 3:15:05 AM UTC+1, Christina Ilvento 
>>>>>> wrote:
>>>>>>>
>>>>>>> Hi All,
>>>>>>>
>>>>>>> Would you mind sending app-ids that you're seeing this behavior for? 
>>>>>>> Please feel free to send them to me directly or to link any issues you 
>>>>>>> have 
>>>>>>> filed in our issue tracker so that we can investigate.
>>>>>>>
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Christina
>>>>>>>
>>>>>>>
>>>>>>> On Fri, Dec 21, 2012 at 8:39 AM, Carl Schroeder <
>>>>>>> [email protected]> wrote:
>>>>>>>
>>>>>>>> I've been sacrificing unicorns to dark powers...clearly I have been 
>>>>>>>> doing it wrong.
>>>>>>>> That probably explains some unusual behavior in other parts of my 
>>>>>>>> app.
>>>>>>>> AFK pentagrams. :(
>>>>>>>>
>>>>>>>>
>>>>>>>> On Thursday, December 20, 2012 7:51:35 PM UTC-8, Cesium wrote:
>>>>>>>>>
>>>>>>>>> Ready for this?
>>>>>>>>>
>>>>>>>>> Now, a single instance survives for hours and hours, happily 
>>>>>>>>> serving requests with the usual low latency response time.
>>>>>>>>>
>>>>>>>>> This is just what Sir Brandon wrote about. Mysterious changes in 
>>>>>>>>> the system's behavior.
>>>>>>>>>
>>>>>>>>> I should note that I sprinkled rainbow Skittles across the floor 
>>>>>>>>> to attract the unicorns. They're back!
>>>>>>>>>
>>>>>>>>> David
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>  -- 
>>>>>>>> You received this message because you are subscribed to the Google 
>>>>>>>> Groups "Google App Engine" group.
>>>>>>>> To view this discussion on the web visit 
>>>>>>>> https://groups.google.com/d/msg/google-appengine/-/MFdFH0ZqWbgJ.
>>>>>>>>
>>>>>>>> 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.
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> -- 
>>>>>>>
>>>>>>> Christina Ilvento | Google App Engine | [email protected] |
>>>>>>>  (650)-201-9399
>>>>>>>
>>>>>>>
>>>>>>> 

-- 
You received this message because you are subscribed to the Google Groups 
"Google App Engine" group.
To view this discussion on the web visit 
https://groups.google.com/d/msg/google-appengine/-/g-XCFMNJHRwJ.
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