Report incorrect username or password

[felipeZ]

If a service is started like

import cerise_client.service as cc

srv = cc.require_managed_service(
        'cerise-mdstudio-das5-myuser', 29593,
        'mdstudio/cerise-mdstudio-das5:develop',
        'user',
        'password')

And there is not user in the remote machine, cerise-client neither complains about it nor finishes the process.

[LourensVeen]

Ah, yes, and we need support for other kinds of credentials as well (#4). Thing is, if you're running locally (and why not have a specialisation for that?), you don't need a username, so we cannot put that check into Cerise Client, because a missing username may actually make sense.

I fully agree that there should be a nice error message. Right now, that error message is in the service log, which is difficult to get out (you have to do a docker cp on the command line). That's #2. Then, we need a useful error message into that log, which is impossible with Xenon 1. I'm in the process of porting Cerise to Xenon 2, which should help, but it's slow going.

Ideally, require_managed_service() would actually throw an exception with a nice error message, so that you don't need to request the log at all.The question is how we communicate the error from Cerise to Cerise Client. Or maybe when Cerise Client starts a service, it should wait a bit, try to connect, and if it can't, throw an exception with the log output in it? In production the log level will be low, so there shouldn't be too much output...

[LourensVeen]

Getting the log we now can do (as of 0.2.2). Now the question is how to get an error message to the user. Perhaps we could have a second log file that only logs e.g. critical-level errors? Then use that as the message if the service does not respond to e.g. a list-jobs request quickly enough? But then again we may as well just get the whole log. Or maybe filter it? Then the user would get the critical error message, and they'd still be able to use get_log() to get everything (to the extent that the specialist has decided to generate it by setting the log level).

[LourensVeen]

As a first order approximation, grepping lines with CRITICAL from the log file, and adding that to the exception message, should give some useful output. Then we need a timeout? And perhaps Cerise should produce a log message when it's ready to accept jobs, which we can scan for to see if it came up alright?