Closes down asyncio server when sending alert email fails

[icr-ctl]

Not well-documented yet but there's a crash occurring that shutdowns the asyncio server (i.e. the server handling connections to the scrubcams). It occurs immediately after a scrubcam sends an image, often its first image. Discovering that I can recreate it almost every (seemingly every) time by showing the Scrubcam the giraffe toy. But when I show it myself (person class): no crash.

[icr-ctl]

When this bug happens SD logs that it is shutting down the asyncio server

[INFO] Successfully shut down asyncio server. (scrubdash.asyncio_server.asyncio_server)

And SC raises a

BrokenPipeError: [Errno 32] Broken pipe

Question is who broke the socket? SD or SC?

I did an ncat -l 8888 on the sagebrush machine and ran scrubcam on eagleogle. scrubcam did not crash which suggests that scrubdash is the one closing the socket. so there are two theories I can think of: SC is sending totally valid stuff and SD has a bug or SD is doing everything fine but SC is sending something bogus which SD is choking on and following up by closing down the asyncio server.

A quick look seems to show that only a KeyboardInterrupt in SD would trigger the shutdown message above (which is curious) but maybe more looking into things is necessary.

[icr-ctl]

Aha! I found where KeyboardInterrupt is raised in the code! And it was being raised inside a caught exception that relates to the SMTP handling. I have not delved too deep but what I think has happened is that Google has blocked the email address we were using to send out alert emails in some way and that is breaking things. The short-term hack is to empty the ALERT_CLASSES: in the config file. Later we can figure out what happened with that email address and how to fix it or what-have-you.

[icr-ctl]

This issue should stay open as a) SD should probably be more verbose about what is going wrong in this failure mode b) SD should perhaps more gracefully handle this scenario and keep the server alive even if Google won't accept the attempt to send the email and c) we should figure out how to keep Google happy about our email-sending or find another way to send the emails.

[iamgarcia]

Email alerts fail
It appears that on the Google account, third-party access was disabled. Enabling third-party access has allowed ScrubDash to successfully send emails to my person email.

Lack of verbosity when email alerts fail
I will make sure to add logging information for when email alerts fail.

Prevent ScrubDash from closing the asyncio server
I need to come up with a better way of handling email alert failures. As of right now, it raises a KeyboardInterrupt. We want the asyncio server to not shut down because the ScrubCam won't be able to send data to the ScrubDash server.