feature: add Anomalist to PROD environment

[paarriagadap]

One-liner

Implement Anomalist in the production enviroment

Context & details

• Currently, Anomalist only works when there is a staging server generated via a PR.
• It would be nice to access some of its checks in PROD, for example when a user asks about weird jumps in the data, or when we are just curious about an indicator previously not reviewed there.
• This could be implemented as part of the checks of the Perfect Data Pages project

Workaround

There is currently a workaround for this, which is opening a new PR, running etl pr "Inspect child mortality anomalies" and play with Anomalist in a staging server.

[pabloarosado]

In my view, it would be uncommon that someone wants to inspect a random dataset, and if so, it would most likely be one of us (data scientists), who can easily use the workaround. I think the workaround should be easy enough, and adjusting Anomalist to properly work in production may cause some headaches. So maybe we can close this issue, but @paarriagadap feel free to reopen it if you disagree or if you think this would be very valuable. Thanks for raising this!

[paarriagadap]

I think that an implementation in production would be useful for researchers as well, in case they have questions about the data. If it's too tricky to implement it's not worth it then. Thanks.

[pabloarosado]

Thanks @paarriagadap for the clarification. My intuition is that researchers would not be working with Anomalist often, and if they wanted to inspect a specific data issue, they could raise it on Slack/GitHub, and we could then help them have a look. But I'll raise this on the pain-points channel to ask if this could be useful for researchers. If so, we can reopen and reassess.

[lucasrodes]

I personally think that having anomalist in PROD could be nice. It would make the tool way more accessible. It would, for instance, remove the middle man (data scientists) so that researchers can go solo and do checks whenever (sometimes someone may work outside of regular hours).

If we don’t want it in PROD bc of it interacting with the DB, we could have a staging server on that is often pulling changes from prod (e.g. staging-site-master).

I don't think we need to discard this idea just yet.

[pabloarosado]

Thanks @lucasrodes, I agree it would be nice. But I also think it won't be used much by researchers (if you disagree, we can ask around to get a better sense), and looking at the packed cycles ahead of us, I don't think we can realistically refactor Anomalist in the coming months. But if you think of a simple solution, feel free to reopen and suggest it.

[lucasrodes]

I agree we are unlikely to tackle this in the coming weeks or months.

But maybe this is something for later on? I wonder where is a good place for this issues, which are not necessarily short-term.

Anyhow, don't feel strong about this. Let's ask and see.