Google Feed API not working today (02/24/2012)

[Carlos Eduardo Benincasa]

For some reason the Google Feed API is not working today.

As you can see in the code playground (http://code.google.com/apis/
ajax/playground/#load_feed), the API is not loading the RSS Feeds at
all.

It worked perfectly since this morning. What happened?

Regards,
Carlos

[Chris]

I have gotten the QPS limit reached repeatedly today.

On Feb 24, 10:53 am, Carlos Eduardo Benincasa <zza...@gmail.com>
wrote:

[Adam Feldman]

Hi Carlos,

Are you still seeing this error?  Everything in the Code Playground sample you mentioned looks okay to me.

Is anyone else also seeing this problem?

Thanks,

Adam

--
You received this message because you are subscribed to the Google
Groups "Google AJAX APIs" group.
To post to this group, send email to
google-ajax...@googlegroups.com
To unsubscribe from this group, send email to
google-ajax-searc...@googlegroups.com
To view this message on the web, visit
http://groups.google.com/group/google-ajax-search-api?hl=en_US
For more options, visit this group at
http://groups.google.com/group/google-ajax-search-api?hl=en?hl=en

[Carlos Eduardo Benincasa]

Hello Adam,

Everything is working fine again!

Maybe it was some service instability, who knows...

Thanks!

[webjockey]

Been having trouble with the QPS limit being reached for a few weeks
now. Feed API worked fine for at least two months prior to this. Have
tried to 'throttle down' the number of items going in with no
noticeable improvement. Here's a couple test pages from my site that
automatically reload every five or ten minutes so you can see what I
mean:

http://www.soonersetc.com/headline-news/
http://www.soonersetc.com/headline-news/index2.html

Interestingly the page that is throttled down the least fails less
often. Don't really understand how Google Code, Feed API, etc. work.
All I know is what used to work great no longer does...

On Feb 24, 10:53 am, Carlos Eduardo Benincasa <zza...@gmail.com>
wrote:

[Jeremy Geerdes]

I have been monitoring the relatively large number of these reports over the last several weeks, but I am finding it puzzling for a number of reasons. The biggest of these are:

First, I have been unable to reproduce the "QPS limit exceeded" error that has been reported. This includes on the pages linked from this and other threads regarding the matter. But also, it includes a couple of attempts I have made utilizing code on my local computer. For example, today, I took the very same code that webjockey used, eliminated anything not related to the Feeds API, and added additional feeds until I was requesting 100 individual feeds. I then refreshed the page 5 times in rapid succession, resulting in 500 individual requests. All of the requests returned as expected.

And second, I don't believe that the Feeds API will ever return an error that reads "QPS limit exceeded." Adam Feldman has supported this belief. If you trigger the throttling mechanisms on the Feeds API, the error message should read "Suspected terms abuse." or something to that effect.

A couple of tidbits that may be related to the issue, but not necessarily:

At least three of the sites reporting issues are using script(s) by Dynamic Drive. I have not taken time to really dissect their code, but there does not seem to be anything unusual. I am not convinced that this is an issue, but it is a commonality, especially considering at least one other site reporting an issue was not (at least when I checked) using Dynamic Drive scripts.

I investigated geographic dispersion of the sites, thinking it may be something with a particular data center. This may also explain why I am unable to reproduce the issue while others report that it persists. But there does not seem to be anything common there. Reports are from sites on at least 3 continents in at least two languages. Although as a non-Googler, I certainly don't have access to their routing and load balancing systems to determine conclusively, I can only guess that there would be multiple data centers involved here.

Given that most or all of the sites reporting issues are running client-side applications, the issue can't be with the particular server(s) hosting the sites. The Feeds API creates a request directly to Google's servers, so once the original page is served, the originating server is not involved in any requests or responses from Google.

All of this adds up to this: we have ourselves a bit of a mystery. The problem is either isolated or intermittent, and without being able to reproduce and analyze it, it's almost impossible to diagnose, let alone fix. If you have any further insights into this issue, please post whatever you possibly can!

Jeremy R. Geerdes
Generally Cool Guy
Des Moines, IA

For more information or a project quote:
jrge...@gmail.com

If you're in the Des Moines, IA, area, check out Debra Heights Wesleyan Church!

[webjockey]

Thanks Jeremy. I'll try and dig up a few more clues as soon as it
'breaks' again. -Blair

On Feb 25, 3:12 pm, Jeremy Geerdes <jrgeer...@gmail.com> wrote:
> I have been monitoring the relatively large number of these reports over the last several weeks, but I am finding it puzzling for a number of reasons. The biggest of these are:
>
> First, I have been unable to reproduce the "QPS limit exceeded" error that has been reported. This includes on the pages linked from this and other threads regarding the matter. But also, it includes a couple of attempts I have made utilizing code on my local computer. For example, today, I took the very same code that webjockey used, eliminated anything not related to the Feeds API, and added additional feeds until I was requesting 100 individual feeds. I then refreshed the page 5 times in rapid succession, resulting in 500 individual requests. All of the requests returned as expected.
>
> And second, I don't believe that the Feeds API will ever return an error that reads "QPS limit exceeded." Adam Feldman has supported this belief. If you trigger the throttling mechanisms on the Feeds API, the error message should read "Suspected terms abuse." or something to that effect.
>
> A couple of tidbits that may be related to the issue, but not necessarily:
>
> At least three of the sites reporting issues are using script(s) by Dynamic Drive. I have not taken time to really dissect their code, but there does not seem to be anything unusual. I am not convinced that this is an issue, but it is a commonality, especially considering at least one other site reporting an issue was not (at least when I checked) using Dynamic Drive scripts.
>
> I investigated geographic dispersion of the sites, thinking it may be something with a particular data center. This may also explain why I am unable to reproduce the issue while others report that it persists. But there does not seem to be anything common there. Reports are from sites on at least 3 continents in at least two languages. Although as a non-Googler, I certainly don't have access to their routing and load balancing systems to determine conclusively, I can only guess that there would be multiple data centers involved here.
>
> Given that most or all of the sites reporting issues are running client-side applications, the issue can't be with the particular server(s) hosting the sites. The Feeds API creates a request directly to Google's servers, so once the original page is served, the originating server is not involved in any requests or responses from Google.
>
> All of this adds up to this: we have ourselves a bit of a mystery. The problem is either isolated or intermittent, and without being able to reproduce and analyze it, it's almost impossible to diagnose, let alone fix. If you have any further insights into this issue, please post whatever you possibly can!
>
> Jeremy R. Geerdes
> Generally Cool Guy
> Des Moines, IA
>
> For more information or a project quote:

> jrgeer...@gmail.com

[omr]

Jeremy Geerdes wrote:
> "... I don't believe that the Feeds API will ever return an error that reads 'QPS limit exceeded.' ..."

Actually the message in question reads: "qps rate exceeeded". Note
the lower-case "qps" and the misspelling: "exceeeded" [sic]. This is
confirmed in other discussions (and a screenshot) posted elsewhere.
The text "qps rate exceeeded" [sic] appears in the "error.message"
property of the result argument that the Feed.load method passes to
its callback.

-- omr

[Jeremy Geerdes]

Thanks for the clarification, omr. I had noticed the spelling error in the threads (one of which carries the error in the subject line) but assumed it was a typo. I have not seen the screenshots you apparently have. Any links would be appreciated, especially if they provide additional information about the queries that are being rejected. At any rate, I have still never seen that error message, and in a brief, off-group conversation with Adam, he indicated that he didn't know that the service would return such an error, either. That's not to say that it won't. Adam is the PM, and I'm just the non-Googler lackey. While I would not presume to speak for him, I suspect that neither of us have the luxury of playing around in the service codebase every moment of the day. So...

We're still stuck on trying to figure out how to reproduce this error. Anyone who can come up with a consistent repro protocol gets a virtual pat on the back from me!

Jeremy R. Geerdes
Generally Cool Guy
Des Moines, IA

For more information or a project quote:
jrge...@gmail.com

If you're in the Des Moines, IA, area, check out Debra Heights Wesleyan Church!

[omr]

Jeremy Geerdes wrote:
> "... Any links would be appreciated, especially if they provide additional information about the queries that are being rejected. ..."

These links won't add much further information, but will confirm what
I already stated.

See the 3rd post (by "sail") in this thread:

http://www.cfd-online.com/Forums/site-news-announcements/97559-feed-aggregations-what-do-you-think.html#post345337

> "... 'Error fetching feeds: qps rate exceeeded' ..."

That discussion is about the reader used at e.g.:

http://www.cfd-online.com/Feeds/blogs.php

The feed script used on that page shows that the "qps rate
exceeeded" [sic] text must have come from the error.message property
of the result object passed to the load method's callback.

http://www.cfd-online.com/Feeds/rssdisplayer.js

Regarding another implementation elsewhere, this screenshot again
confirms the "exceeeded" misspelling in the error message text:

http://goo.gl/46d0d

(The image is too small, zoom in to see the text.)

Here is another report, from Japan:

http://detail.chiebukuro.yahoo.co.jp/qa/question_detail/q1381937188

-- omr

[Blair Shorney]

Thanks omr. This is precisely what I was getting:

http://img84.imageshack.us/img84/6748/90835727.jpg

Dimmed out screen with the pop-up. Would have to press the 'OK' button multiple times (assume it was a stack of them) and the screen would slowly lighten up until it displayed a whole string of something like:

Undefined
invalid date

Presumably one for each feed item.

Anyway, still waiting for a recurrence. 

-- omr

[omr]

BTW:

In October 2010 a user observed the "qps rate exceeeded" [sic] error
message in Maps API geocoding responses:

http://groups.google.com/group/google-ajax-search-api/browse_thread/thread/f012c6b4db2095b3/26629098e00e065c

-- omr

[Jeremy Geerdes]

Yeah, I heard it could come from the geocoder with Maps. Thanks for the additional info. That screen shot looks like a Dynamic Drive script. They do a simple alert with "Google Feeds API error:" followed by the error.message. So you're probably right that it's coming from the API at some point (i.e., Adam and I were mistaken). Even so, the inability to reproduce the issue with 100 feeds on a page, refreshed 5 times in rapid succession is puzzling. Still looking for that repro protocol.

Jg

[webjockey]

Here we go folks:

http://www.soonersetc.com/google-api-error-01.jpg

After clicking 'OK' 5 times:

http://www.soonersetc.com/google-api-error-02.jpg

After clicking 'OK' 10 times:

http://www.soonersetc.com/google-api-error-03.jpg

After clicking again feed items displayed normally. Then re-loaded
page and got nothing but 'Retrieving RSS feed(s)'. Still not working
at time of this post.

On Feb 25, 9:40 pm, omr <omr99...@yahoo.com> wrote:
> Jeremy Geerdes wrote:
> > "... Any links would be appreciated, especially if they provide additional information about the queries that are being rejected. ..."
>
> These links won't add much further information, but will confirm what
> I already stated.
>
> See the 3rd post (by "sail") in this thread:
>

>  http://www.cfd-online.com/Forums/site-news-announcements/97559-feed-a...

[Jeremy Geerdes]

Okay, so you're still seeing it. How did you get it to throw the error the first time? Because I just went to the page referenced in your screenshots and refreshed 10 times without seeing an error. Is the error still happening. If so, can you provide step-by-step instructions of how you did it?

Jeremy R. Geerdes
Generally Cool Guy
Des Moines, IA

For more information or a project quote:
jrge...@gmail.com

If you're in the Des Moines, IA, area, check out Debra Heights Wesleyan Church!

[Adam Feldman]

Hi folks,

Thanks for the extra information, it's very useful.  I think I know where to look now - give me some time to investigate, and I'll let you know what we find.

Cheers,

Adam

[webjockey]

I just leave that test page with the meta refresh running on one of my
tabs and check it from time to time:

http://www.soonersetc.com/headline-news/index2.html

This morning was the first and only time since my original post on
this thread that I got the error. Don't know of any way to get it to
break. Once it does break it'll usually come back after clearing the
pop-ups and reloading the page. Then sometimes get this:

undefined [undefined] Invalid Date
undefined

undefined [undefined] Invalid Date
undefined

Which if you click on will send you to a 404 page with a url like
these:

http://www.soonersetc.com/headline-news/function%20link()%20%7B%20%20%20%20%5Bnative%20code%5D%7D

http://www.soonersetc.com/headline-news/function%20link%28%29%20%7B%20%20%20%20%5Bnative%20code%5D%7D

Error is very intermittent, once or twice a day, sometimes not at all.

On Feb 26, 3:13 pm, Jeremy Geerdes <jrgeer...@gmail.com> wrote:
> Okay, so you're still seeing it. How did you get it to throw the error the first time? Because I just went to the page referenced in your screenshots and refreshed 10 times without seeing an error. Is the error still happening. If so, can you provide step-by-step instructions of how you did it?
>
> Jeremy R. Geerdes
> Generally Cool Guy
> Des Moines, IA
>
> For more information or a project quote:

> jrgeer...@gmail.com

[Adam Feldman]

Hi everyone,

Thanks again for the help.  We've found the problem and are working on a fix.  I'll report back when it's taken care of.

Cheers,

Adam

[webjockey]

Thanks Adam.

On Feb 27, 1:17 pm, Adam Feldman <adam.feld...@google.com> wrote:
> Hi everyone,
>
> Thanks again for the help.  We've found the problem and are working on a
> fix.  I'll report back when it's taken care of.
>
> Cheers,
> Adam
>
>
>
>
>
>
>
> On Sun, Feb 26, 2012 at 2:05 PM, webjockey <bashorn...@gmail.com> wrote:
> > I just leave that test page with the meta refresh running on one of my
> > tabs and check it from time to time:
>
> >http://www.soonersetc.com/headline-news/index2.html
>
> > This morning was the first and only time since my original post on
> > this thread that I got the error. Don't know of any way to get it to
> > break. Once it does break it'll usually come back after clearing the
> > pop-ups and reloading the page. Then sometimes get this:
>
> > undefined [undefined] Invalid Date
> > undefined
>
> > undefined [undefined] Invalid Date
> > undefined
>
> > Which if you click on will send you to a 404 page with a url like
> > these:
>

> >http://www.soonersetc.com/headline-news/function%20link()%20%7B%20%20...
>
> >http://www.soonersetc.com/headline-news/function%20link%28%29%20%7B%2...

[Adam Feldman]

Hi folks,

Sorry for the delay.  This was a global issue, which explains why no individual sites got the message consistently.  It shouldn't reoccur.  If it does, please ping this thread.

Thanks to everyone who helped hunt this down, especially Jeremy and omr!

Thanks,

Adam

[webjockey]

A belated thanks to Jeremy and omr. Thanked Adam but forgot you two.
Really needed this issue resolved by today and it was. I appreciate
all of your help in this matter. -Blair