[Labs-l] Site monitoring shown "Site down" every day during 05:20 to 05:30 UTC

Fri May 9 00:36:47 UTC 2014

Rainer Rillke <rainerrillke at hotmail.com> wrote:

> I am sorry for not yet being able to provide you with details but
> monitoring my webservices from external servers, I found that at least
> one of my tool's sites is considered down between 05:20 to 05:30 UTC
> every day.

> Is there some heavy tool running on this time? Maintenance window? Known
> issue?

> During the request, neither database connection, nor PHP is tested; just
> whether the webservice responds with a non-empty 200 OK result.

> The tools that are reported to be down are rillke and expose-data.

expose-data has no ~/{access,error}.log (did you delete them
/after/ starting the webservice?) so it's hard to tell.

Much harder to understand though is rillke's behaviour: Ac-
cording to ~/access.log, it should be "down" by the defini-
tion above most of the time as it returns 304s ("Not modi-
fied"), and that also around 5:20Z.  The logs on
tools-webproxy don't look that different, and a /very/ cur-
sory look at the number of requests handled suggests that
5:20Z to 5:30Z is no peak.

So: Not a clue.

Tim