Fetch current OSD pool configuration over the API

[masnax]

PUT /1.0/pools-op lets you set the replication size for several OSD pools.

This PR adds the ability to fetch the current sizes (and some other relevant OSD pool information) via GET /1.0/pools. The goal of this is to obtain the correct OSD pool names to supply to PUT /1.0/pools-op, as well as to determine what the current size of each pool is.

As discussed on mattermost: https://chat.canonical.com/canonical/pl/g3heo1kmz7r6db39re9qyzfw4w

[UtkarshBhatthere]

LGTM, one minor nit.

[masnax]

@UtkarshBhatthere @sabaini @lmlg Sort of a side note/question here: I've been thinking about the PUT /1.0/pools-op implementation a bit and I'm wondering if it's a bit aggressive?

Right now it takes 1 size value, and applies it to all specified OSD pools, but also sets the default OSD pool size for all future pools. This means if you want to resize just 1 pool, you have to call this API twice, because the first call will update the default OSD pool size as well, so you'd have to restore that with another request but with a blank set of OSD pools.

I wonder if we can make the default OSD pool size update based on a query parameter? Like ?set-default=1 so that you're not locked into changing the default pool size. To accommodate the query param in the CLI, we can add a --set-default flag too, to set the default OSD pool size.

I noticed the CLI help is also wrong. It says microceph pool set-rf <SIZE> <POOLS>... but the only arguments it interprets are <POOLS>... with size being supplied by --size.

[UtkarshBhatthere]

@masnax fair enough, can you please create a GH issue for this behaviour ? We will take it up soon.

[sabaini]

Thanks @masnax , this lgtm. It would be great to have a test for this though

[UtkarshBhatthere]

One change requested

[lmlg]

@UtkarshBhatthere @sabaini @lmlg Sort of a side note/question here: I've been thinking about the PUT /1.0/pools-op implementation a bit and I'm wondering if it's a bit aggressive?

Right now it takes 1 size value, and applies it to all specified OSD pools, but also sets the default OSD pool size for all future pools. This means if you want to resize just 1 pool, you have to call this API twice, because the first call will update the default OSD pool size as well, so you'd have to restore that with another request but with a blank set of OSD pools.

I wonder if we can make the default OSD pool size update based on a query parameter? Like ?set-default=1 so that you're not locked into changing the default pool size. To accommodate the query param in the CLI, we can add a --set-default flag too, to set the default OSD pool size.

I noticed the CLI help is also wrong. It says microceph pool set-rf <SIZE> <POOLS>... but the only arguments it interprets are <POOLS>... with size being supplied by --size.

Yes, those look line sane changes to me. Not sure if there's any client that requires the current command format to be stable, though.

[UtkarshBhatthere]

Fair enough @lmlg, I will share a post in Ceph Announcements room with the newly created ticket and the description of the change.