Solve issue #42 (AIOpsLab)

[ChuanweiQu]

Change the setting of hotel-reservation application to use "standard" storageClass to provision PVs dynamically. In this way, when PVCs are deleted, PVs will be cleaned up automatically.

[ChuanweiQu]

@HacksonClark Hi, Jackson. I have already changed the pvc setting for hotel-reservation and this works well. Do you need to review my changes?

[yinfangchen]

Good catch @ChuanweiQu! However, I think this may break a fault that relies on this behavior - redeploy_without_PV-
Basically, every time the PVC is deleted, the PV will persist there; and next time when you re-deploy the application, there will be errors happening. This is how the fault happens for this scenario.

[HacksonClark]

@yinfangchen That's true, maybe we need a more creative solution.

[ChuanweiQu]

Okay. I will try to find another solution and see if it works.

[ChuanweiQu]

I've been thinking about this carefully. The source of exceptions that we can not control should only be the the call of agent (such as network issues or authority issues.) If exceptions happen in other parts of our code, that means we have bugs and we should fix the bugs to prevent exceptions instead of thinking about exit cleanly. So, the only situation we care about is when the exception happen during call of agent. In this way, it won't be hard to apply try catch statement.

[HacksonClark]

@ChuanweiQu That makes sense to me. There's only so much we can do about a user running Ctrl+C.

[ChuanweiQu]

@HacksonClark No, we can still use atexit to deal with the situation of ctrl+c. I'm just saying if users try to catch some exceptions and not exit, the function registered with atexit will not work. However, If we catch the expcetions first in our code, this will not be a problem. And as I mentioned above, the only source of exceptions should be the call of agents.

[HacksonClark]

@ChuanweiQu I see. This seems like a good approach.