cache timeout on run-action command

[Tom Haddon]

I have a machine model running 2.9.42 on OpenStack, and have received the following output from a `run-action` command:

```
$ juju run-action large/1 --wait check-runners
{}
ERROR model cache: model "bc8ac420-20d2-4648-89d7-c42c7fe4d256" did not appear in cache timeout
```

I'm not sure what this means or how to deal with it.

[Tom Haddon (mthaddon)]

Before getting this particular error I'd run this:

```
$ juju run-action large/1 large/3 large/5 large/6 large/11 large/13 --wait check-runners
{}
ERROR connection is shut down
```

I then tried iterating through the units to see if there was one with a particular problem. The others all returned the output of `run-action` without an issue, seems whatever the problem is it only affects large/1.

[Joseph Phillips (manadart)]

Can we get the controller logs for this one?

It usually alludes to a data integrity problem, where the cache can not be primed with the model's data. This prevents the particular model being used by the server-side API.

[Tom Haddon (mthaddon)]

(Sorry, Canonical only visibility on these).

https://pastebin.canonical.com/p/FqH6GMMzs3/ are the unit logs and https://pastebin.canonical.com/p/HNYMJP5ZGW/ are the controller logs.

[Ian Booth (wallyworld)]

The model cache is an unfortunate artifact which doesn't carry its weight - it is expensive to set up and keep in sync. In fact, it is currently only used to read app/unit status values for the juju status command, and provide logging config for a model. Both of these can be done by querying the model in the db directly.

In the main branch, we have removed it entirely. I wonder if we should backport this removal to 2.9.