ReportDownloadError.ERROR_GETTING_RESPONSE_FROM_BACKEND

[Sandy B.]

Hi All,

since the beginning of april we have a problem to get CRITERIONCREATIVE reports for impression and clicks for some accounts.

We tried to get reports for a smaler size of campaigns for an account (just active campaigns). That seems to work. But the whole account failed with:

Exception fetching Stats - Report download failed. Underlying errors are
Type = 'ReportDownloadError.ERROR_GETTING_RESPONSE_FROM_BACKEND', Trigger = 'Unable to read report data', FieldPath = ''.

Because the reports are downloaded by curl we cant send a SOAP request for further information. Just these params:

<?xml version="1.0"?>
<reportDefinition>
  <selector>
    <fields>Date</fields>
    <fields>CampaignId</fields>
    <fields>AdGroupId</fields>
    <fields>KeywordId</fields>
    <fields>Id</fields>
    <fields>Cost</fields>
    <fields>Impressions</fields>
    <fields>AveragePosition</fields>
    <predicates>
      <field>Status</field>
      <operator>IN</operator>
      <values>ENABLED</values>
      <values>PAUSED</values>
      <values>DISABLED</values>
    </predicates>
    <dateRange>
      <min>20150408</min>
      <max>20150409</max>
    </dateRange>
  </selector>
  <reportName>Custom TYPE_REPORT_CRITERIONCREATIVE AGGREGATION_DAILY METRIC_IMPRESSIONS report #1429015137</reportName>
  <reportType>AD_PERFORMANCE_REPORT</reportType>
  <dateRangeType>CUSTOM_DATE</dateRangeType>
  <downloadFormat>GZIPPED_CSV</downloadFormat>
</reportDefinition>

I've read in other threads that the problem is a kind of runtime problem. So we tried a date range by one as smallest too. but it doesnt work nevertheless.

Is there any solution for our problem?

[Danial Klimkin]

Hello Sandy,

What is the clientCustomerId of the account this report fails for?

Also, note this definition covers two days, I suggest trying with a single day (min == max in dateRange).

-Danial, AdWords API Team.

[Sandy B.]

Hi Danial,

thx for your response. 

The clientCostumerId for this account is '8767143936' (but some other accounts are affected too).

Sry for the confusion with the daterange. i forgot to write, that our script is splitting the daterange after every failure till it tries a single date to get. The params xml was the first one which fails. 

The cron job last night failed again. But i tried it some minutes ago again and it worked for diffrent date ranges as well as for a single date. 

We cant explain it. Maybe its possible that we stress your system, when we try to get the reports of all our accounts at the same time and actually nothing is running simultaneously. But that would not explain while it failed yesterday in the afternoon too.

[Danial Klimkin]

Hello Sandy,

I ran the report for this account a few times at different daytime and it worked well -- returned in seconds.

Do you see this issue for this particular account regularly, or is sporadic? Can you send me a few recent examples of CID - timestamp (no older than 24 hours)?

How exactly do you run reports, is it a single or multiple threads?

-Danial, AdWords API Team.

[Sandy B.]

Hi Danial,

i checked the nightly failure mails and our problem account failed again at 1:44:16.

some other accounts and times in the last 24 h which failed:

23:37:18 - 3525325529

22:44:12 - 1838456912

18:22:46 - 4489247030

17:05:45 - 1773762480

14:53:37 - 6365852806

13:44:52 - 4854490525

13:25:12 - 7866534376

And how it works in our system:

Starting at 01:15h GMT + 2 we download statistics for all our google accounts (round about 30k). Those requests are bunched, so a maximum of 300 curl requests during the same time can be active from 20 different ip addresses within our range.

I hope  these information will help a bit more to solve the problem. 

Thank your for your cooperation :)

[Danial Klimkin]

Hello Sandy,

Which timezone are these timestamps in? GMT+2?

300 concurrent requests sounds a bit extreme. How long does the overall process takes? Can you consider lowering the concurrency level?

We'll if anything can be improved from our side, but generally we expect some low number of failures for this number of requests, which should be fully resolved with a retry. Do you put an account to the end of the queue if a request fails?

-Danial, AdWords API Team.

[Sandy B.]

Hi Danial,

it seems that last night nothing failed. I hope its a good sign. The timezone  I missed is GMT+2.

We will check if we can reduce the 300 requests in any way but we see less chances for it. 

Actually we do the retry direct after the failed request and put it not again at the end of the queue. We will check if it would help, when we changes that behavior.

So right now we will just observer it for a while if the problem will occur again. Normally its no problem for us, if some request fail, because we will get the data with the retry or one day later, but if the same account always fails, then we have a problem, because 

 we dont get the neccessary data at all.

[Danial Klimkin]

Thanks for the details. Please let us know if re-ordering helped.

-Danial.