[OSM-dev] upload josm files from console
dave at sr71.net
Tue Sep 4 22:09:10 BST 2007
On Tue, 2007-09-04 at 22:59 +0200, Frederik Ramm wrote:
> I see both your points. Imi is right in that you should not normally
> retry after a 500 error; but then again it works in practice - as the
> API sometimes gives us a 500 error when in fact some internal timeout
> broke things and a retry will very well work. And everybody says OSM
> is ruthlessly pragmatic!
I asked on dev@ how to handle these properly. According to TomH: 'A 500
error doesn't mean "server too busy"'. I've attached his reply.
Although Tom fixed the server to more aggressively retry when it sees
500 errors due to mysql deadlocks, I'm still seeing a few of them. So,
this isn't as big of a deal as it was a week ago.
> About retrying in case of timeouts - I think it is quite usual to do
> so, at least if you use exponential backoff. Maybe we should ask Tom
> Hughes what he thinks about this since he has to mend the server
> afterwards ;-)
Yeah, timeouts can happen for a whole host of reasons, not just server
congenstion. NAT boxes get rebooted, wireless networks glitch, etc...
Frankly, if the network times out (and had a reasonable timeout to begin
with), then you've probably waited long enough by the time the timeout
-------------- next part --------------
An embedded message was scrubbed...
From: Tom Hughes <tom at compton.nu>
Subject: Re: [OSM-dev] Handling 500 errors
Date: Fri, 31 Aug 2007 00:36:15 +0100
More information about the dev