exceptions in PeriodicImageDownloadService will cause it to stop running

[Blake Rouse]

Exceptions from _start_download() -- or anywhere in this function -- will propagate up to the TimerService (well, the LoopingCall that it uses internally) and cause it to stop running; it considers any uncaught error as fatal. You ought to catch exceptions and log them instead of letting them get up to the service. I've used a wrapper function named with the prefix try_ in lp:~allenap/maas/safe-check-node-power to do this.

[Julian Edwards (julian-edwards)]

What exceptions was it raising, out of interest?

[Blake Rouse (blake-rouse)]

It was not raising one specifically, but I have seen it raise one about the gpg key. I dont remember specifically what the name of the exception was.

I think Gavin was just talking in general, it doesn't matter what exception, it will fail. Here is the review that this bug is referencing.

https://code.launchpad.net/~blake-rouse/maas/rpc-import-images/+merge/232271

I think it would be nice to have a exception handling TimerService base class that just logs exceptions and keeps the timer service running.

[Julian Edwards (julian-edwards)]

On Wednesday 27 August 2014 00:31:20 you wrote:
> It was not raising one specifically, but I have seen it raise one about
> the gpg key. I dont remember specifically what the name of the exception
> was.
>
> I think Gavin was just talking in general, it doesn't matter what
> exception, it will fail. Here is the review that this bug is
> referencing.
>
> https://code.launchpad.net/~blake-rouse/maas/rpc-import-> images/+merge/232271
>
> I think it would be nice to have a exception handling TimerService base
> class that just logs exceptions and keeps the timer service running.

Yeah agreed, thanks.

What we need is a base class that:

1. handles connection retries
2. other boilerplate for timer-based services that use RPC
3. handles errors and logs them without taking out the service loop