Sometimes .appspot.com just doesn't respond

[Kaan Soral]

This was an issue that I've been hunting for years, some people report "white pages" on my old app system (facebook apps) - I never could reproduced the issue, I assumed it was a rare client-side issue

Now, while testing my app, I've noticed the same issue, .appspot.com of my app just wasn't responding, I was only able to get it to respond my refreshing several times

The app is simple, there are no errors logged, the request simply doesn't go through

Never noticed the same issue on another app, yet since this other app is mostly Cloudflare wrapped, I wouldn't notice it

Any ideas?

It's just sad knowing this issue exists in the wild, yet I have no way of fixing it, it's causing me to lose users slowly

[Anastasios Hatzis]

If you inspect the "white pages" response in e.g. Chrome dev-tools or Firebug etc., is the response really empty or does it have broken markup or something? What does the response header's content-length say?

Maybe you could add a logging to your app for any such response and then see, if this is really an issue of the front-end servers, or if there is some strange edge case in your app that occasionally produces such responses.

--
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-appengi...@googlegroups.com.
To post to this group, send email to google-a...@googlegroups.com.
Visit this group at https://groups.google.com/group/google-appengine.
To view this discussion on the web visit https://groups.google.com/d/msgid/google-appengine/b508fc29-54e8-4f2e-b59f-e6b049214006%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Mit freundlichen Grüßen / Kind regards

i. A.
Anastasios Hatzis

Fon: +49 8374 930813
Fax: +49 8374 930810
Mobil: +49 1520 8592878

HATZIS Edelstahlbearbeitung GmbH
Hojen 2
87490 Haldenwang (Allgäu)
Germany

Handelsregister Kempten (Allgäu): HRB 4204
Geschäftsführer: Paulos Hatzis, Charalampos Hatzis
Umsatzsteuer-Identifikationsnummer: DE 128791802
GLN: 42 504331 0000 6

http://www.hatzis.de/

[Nicholas (Google Cloud Support)]

I concur with Anastasios. I would suggest inspecting the actual network response using Chrome's Developer Tools the Firebug extension for Firefox. A white page does not necessarily indicate the request didn't go through. In fact, requests not going through often result in standard browser error messages. A plain white page without content can easily be a text/plain response of 0 length or a text/html response with non-conforming markup.

The network request may reveal a more information clue about how those white pages are received.

[Kaan Soral]

Like I mentioned in my OP, the white page is the result of a lingering request with no response, this much I deducted from the browser debuggers

Yet, the issue hit me when I was deep in another task, so I couldn't/didn't trace the network issue beyond the simple browser check

(I regret not debugging it further when I had the issue producing)

I'm 99.9% sure it's not a code issue, the app is very old and robust, the requests in question are 200-300ms

In my opinion, it's either:

1) Instance issue, no instances are assigned to requests

2) Local/temporary network issue, the requests are coming from Turkey from the main ISP

(1) seems more likely to me, these also happened after a deployment, and seemed to go away after another deployment, so I might have caught a better cluster during my deployments

-- -- --

Actually, not reflecting on the issue, (2) seems very likely too, because regularly, all google.com services are unreachable from Turkey for 2-3 minutes, I sometimes can't reach gmail, the issue is fixes itself after 2-3 minutes

[Nick]

I have seen 'outages' like this when deploying over the default instance in the past - a small window of about 30 seconds where no content was served.

When this was happening we worked around it by deploying to another version-id  and switching the default after the first instance spun up.

[Wolfram Gürlich]

Hi Kaan,

you're not alone. We noticed the same issue, too. We have server processes making like 20request/second all the time to appengine, and all of the sudden we were experiencing these lingering requests. It all began earlier this year and still continues. Those request don't leave any traces in the logs and keep the TCP connection open very long (like at least 20 minutes). They never give any reponse at all. 

The funny thing is, we first encountered these request while connecting from a GCE instance. This was over .appspot.com. The error was gone when we switched the connection to a custom domain. But that didn't last. We then saw this errors on other machines, too - not only from GCE. And they also appeared on the custom domain (SNI SSL) and also over VIP SSL.

So I would think it's neither 1) or 2). Instead I would think it is a problem with the Google frontend server instances that get somehow stuck when dispatching the requests.

[Kaan Soral]

Thanks a lot for the reply Wolfram, I hope it urges the App Engine/Google team to give this issue a proper inspection

My issue was exactly like yours, the connection stays open, browser lingers, no logs whatsoever from the App Engine side

Over the course of 3 years, I received various reports of "white pages" - it's how users interpret the issue, since these un-answered but lingering requests result in browser "white pages" - so I assume the issue we experienced and this user-issue might be linked, so it might span 1+ years, however they might be unrelated too

I really hate that issues like these cause a product to decay, I probably lost a lot of users to this "white page" issue, because only a small fraction of my userbase use the report function, maybe 0.01%, yet I received countless reports

[Nicholas (Google Cloud Support)]

Thank you for posting further details here.

I'm not aware of any known issues related to network connectivity to/from Turkey. Nevertheless, network diagnostic tools such as dig and tracert could be helpful in gathering more evidence for this being the issue.

As for this issue pertaining specifically to App Engine, we would really need some evidence supporting a reproducible event not behaving as intended. Unfortunately, transient issues are very difficult to diagnose and investigate. If you have a code sample that can reliably reproduce the lingering requests you've encountered or a precise time frame during which regional connectivity was problematic, feel free to post it here and we will do our best to investigate.

On Wednesday, April 6, 2016 at 4:21:37 PM UTC-4, Kaan Soral wrote: