Lots of DownloadError exceptions
Deepak Sarda
One of my apps is getting DownloadError (ApplicationError 5) for
pretty much all urlfetch remote url fetch requests to a particular
remote server for the past week. Fetch requests to other servers are
not affected. Also, there are no problems fetching content from this
remote server from the local development server. Here's a traceback
from the logs:
Error fetching url
http://www.sbstransit.com.sg/mobileiris/index_svclist.aspx?stopcode=17069
Traceback (most recent call last):
File "/base/data/home/apps/sbsnextbus/6.332480362470588654/utils.py",
line 71, in get_url
result = urlfetch.fetch(url)
File "/base/python_lib/versions/1/google/appengine/api/urlfetch.py",
line 273, in fetch
raise DownloadError(str(e))
DownloadError: ApplicationError: 5
As can be seen from the trace, my app (http://sbsnextbus.appspot.com)
is trying to fetch pages from sbstransit.com.sg
This appears to be a deadline exceeded error (error code 5) but I
can't figure out why this should occur. Everything was running
smoothly till last week when this started happening.
Since I can't reproduce this problem locally and there is no problem
fetching pages from other remote servers in the production
environment, I suspect the issue is related to network connectivity
issues between AppEngine data centers & the servers of
www.sbstransit.com.sg
This problem has been around for almost a week now and I can't seem to
figure out a solution or work-around. Searches in the group archive
didn't yield anything helpful. Any help would be greatly appreciated!
Thanks,
Deepak
PS: Is there a process to open a 'support ticket' with appengine team
regarding infrastructure issues?
Jeff S
The error you are seeing usually means that the urlfetch service did
not receive the reply within ~5 seconds. There could be several
possible sourced of latency which could be involved as you noted, so
do you think the issue could be that the sbstransit site is responding
more slowly than it used to?
Thank you,
Jeff
On Apr 6, 5:45 am, Deepak Sarda <deepak.sa...@gmail.com> wrote:
> Hi
>
> One of my apps is getting DownloadError (ApplicationError 5) for
> pretty much all urlfetch remote url fetch requests to a particular
> remote server for the past week. Fetch requests to other servers are
> not affected. Also, there are no problems fetching content from this
> remote server from the local development server. Here's a traceback
> from the logs:
>
Deepak Sarda
Thanks for your reply. I did some measurements.
First is a request made to a different remote host:
LTA Fetch on App Engine
===================
#
04-06 04:35PM 03.011 /api/v1/16061/30/ 200 768ms 168ms-cpu 0kb Mozilla/
5.0 (X11; U; Linux i686; en-US; rv:1.9.1b3) Gecko/20090305 Firefox/
3.1b3,gzip(gfe)
See details 218.186.10.253 - - [06/Apr/2009:16:35:03 -0700] "GET /api/
v1/16061/30/ HTTP/1.1" 200 275 - "Mozilla/5.0 (X11; U; Linux i686; en-
US; rv:1.9.1b3) Gecko/20090305 Firefox/3.1b3,gzip(gfe)"
#
I 04-06 04:35PM 03.027
Method get called for APIEndPoint
#
D 04-06 04:35PM 03.036
Cache hit for stop 16061
#
I 04-06 04:35PM 03.043
Fetching URL:
http://www.publictransport.sg/public/ptp/en/Getting-Around/ArrivaltimeResults.html?hidServiceNoValue=&hidBusStopValue=16061
#
I 04-06 04:35PM 03.626
Fetched URL:
http://www.publictransport.sg/public/ptp/en/Getting-Around/ArrivaltimeResults.html?hidServiceNoValue=&hidBusStopValue=16061
=========
As you can see, this urlfetch took just over half a second to
complete.
Now a request that caused a fetch from sbstransit.com.sg
NextBus fetch on App Engine
======================
#
04-06 04:34PM 11.864 /api/v1/17069/030/ 200 3276ms 279ms-cpu 0kb
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.1b3) Gecko/20090305
Firefox/3.1b3,gzip(gfe)
See details 218.186.10.253 - - [06/Apr/2009:16:34:15 -0700] "GET /api/
v1/17069/030/ HTTP/1.1" 200 173 - "Mozilla/5.0 (X11; U; Linux i686; en-
US; rv:1.9.1b3) Gecko/20090305 Firefox/3.1b3,gzip(gfe)"
#
I 04-06 04:34PM 12.073
Method get called for APIEndPoint
#
I 04-06 04:34PM 12.107
Fetching URL: http://www.sbstransit.com.sg/mobileiris/index_svclist.aspx?stopcode=17069
#
W 04-06 04:34PM 15.120
utils.py", line 72, in get_url
File "/base/python_lib/versions/1/google/appengine/api/
urlfetch.py", line 273, in fetch
raise DownloadError(str(e))
DownloadError: ApplicationError: 5
Now, the urlfetch got cut-off in 3 seconds. This is longer than usual
but still under the 5 second limit. Now for comparison, here is a log
of the same request executed on my localhost in the development
server.
On localhost
==========
INFO 2009-04-06 23:50:15,961 main.py] Method get called for
APIEndPoint
DEBUG 2009-04-06 23:50:15,962 nextbus.py] Cache hit for stop 17069
INFO 2009-04-06 23:50:15,962 utils.py] Fetching URL:
http://www.sbstransit.com.sg/mobileiris/index_mobresult.aspx?stop=17069&svc=051
DEBUG 2009-04-06 23:50:15,964 urlfetch_stub.py] Making HTTP
request: host = www.sbstransit.com.sg, url =
http://www.sbstransit.com.sg/mobileiris/index_mobresult.aspx?stop=17069&svc=051,
payload = None, headers = {'Host': 'www.sbstransit.com.sg', 'Accept':
'*/*'}
DEBUG 2009-04-06 23:50:16,077 urlfetch_stub.py] Making HTTP
request: host = www.sbstransit.com.sg, url =
http://www.sbstransit.com.sg/mobileiris/(yvyrjnatocvc0y55rjqlocu3)/index_mobresult.aspx?stop=17069&svc=051,
payload = None, headers = {'Host': 'www.sbstransit.com.sg', 'Accept':
'*/*'}
INFO 2009-04-06 23:50:17,259 utils.py] Fetched URL:
http://www.sbstransit.com.sg/mobileiris/index_mobresult.aspx?stop=17069&svc=051
DEBUG 2009-04-06 23:50:17,260 nextbus.py] Saved cache for stop, svc
17069, 051
INFO 2009-04-06 23:50:17,274 dev_appserver.py] "GET /api/
v1/17069/051/ HTTP/1.1" 200 -
=============
On localhost, the urlfetch completed in just under 1.5 seconds.
Doesn't this indicate that (a) this could be a latency issue between
Google's servers & SBS Transit's servers? and (b) request should still
not be cut-off since it is well under the ~5 seconds limit.
I know the 5 seconds limit is not documented (at least I can't find
it) but it is an issue if it keeps changing all the time.
Any thoughts?
Cheers,
Deepak