Possible issue with the Runtime Tutorial

[samuel-denton]

I am unsure what has changed, but both me and Scott are getting errors when running this tutorial.

Log
Tutorial

The error suggests an environment variable was previously being treated as a string but is now acting like an int.
Looking at the rest of the code, that env var ['SITE_ID'] is used as a string in a few places, and cast to an int explicitly in others.

I'm not sure where the tutorials are hosted so I can't see if anything has changed recently. The fix is simple but I would like to understand what changed.

[samuel-denton]

I think this line causes the issue:

This function takes site_id and finds the nearest actual site, then returns the site_id as a int. Not sure if this was recently added or if the order of main has changed. Removing that int cast results in further errors related to the file not being found.

Without knowing how that tutorial is meant to run, its hard to say, but it appears that the problematic function is actually a fallback that shouldn't be used by default, so it's possible it's been broken for a while but not been used. It seems to rely on a data dir existing in the workflow which does not.

As for why it is now trying to use the fallback; the data being returned from www.ogimet.com seems to be a html page saying the request failed so I guess something changed at their end or we are being rate limited?

[oliver-sanders]

    [os.environ['CYLC_WORKFLOW_RUN_DIR'], 'data', time, site_id + '.json']
                                                        ~~~~~~~~^~~~~~~~~
TypeError: unsupported operand type(s) for +: 'int' and 'str'

Try changing that to str(site_id), or changing to a format string.

[oliver-sanders]

As for why it is now trying to use the fallback; the data being returned from www.ogimet.com seems to be a html page saying the request failed so I guess something changed at their end or we are being rate limited?

The error is "Failed to connect to server.", so this isn't rate-limiting, the site is just down for some reason?

[oliver-sanders]

Hmmm, I can access https://www.ogimet.com/home.phtml.en from some browsers, but not others, something funny is going on with this.

[samuel-denton]

I'm similarly getting inconsistent returns from ogimet.com, hopefully just a temporary issue at their end.

I will look at making the error handling and fallback functions a bit more robust and fix the type casting.
And add some logging so a). it's clear which data is used and b). if the fallback is triggered, why