cli/demo: pre-load 'movr' by default

[knz]

Requested/suggested by @kenliu.
cc @piyush-singh

Release note (cli change): cockroach demo without argument becomes
equivalent to cockroach demo movr. The previous behavior
(no pre-defined dataset) is still available via cockroach demo --empty.

[cockroach-teamcity]

This change is 

[jordanlewis]

LGTM. You might want to add some more text to the intro paragraph of cockroach demo now. It currently says:

#
# Welcome to the CockroachDB demo database!
#
# You are connected to a temporary, in-memory CockroachDB
# cluster of 3 nodes. Your changes will not be saved!
#
# Web UI: http://127.0.0.1:57774
#
# Server version: CockroachDB CCL unknown (darwin amd64, built , go1.12.4) (same version as client)
# Cluster ID: 66a60ca1-d154-48df-bb99-7a779da9d08d
#
# Enter \? for a brief introduction.
#

Perhaps in the first paragraph you could say that it's been preloaded with a fictional ridesharing dataset called movr?

[knz]

Perhaps in the first paragraph you could say that it's been preloaded with a fictional ridesharing dataset called movr?

Good idea. Done.

TFYR!

bors r+

[craig]

Build succeeded

• GitHub CI (Cockroach)

[kenliu]

thanks @knz !

closes #39156

[danhhz]

Awesome!

[awoods187]

Should we backport this?

[knz]

It's a feature change, so as per policy... no?

[danhhz]

For the same reason, the movr workload itself has not been backported. I've asked the docs team to ping me if it ends up being a big pain for them that movr is not in 19.1 (and that I'd then go feel out if we'd be willing to make an exception), but so far I haven't heard anything on that.

[jseldess]

If we can make an exception on this, I'd love to get the movr workload (and demo movr) into 19.1, given that customers tend to stay on the stable release for a long while before upgrading, and we're starting to leverage movr throughout our SQL docs.

[danhhz]

Well that's pretty definitive as far as interest goes. @bdarnell are we comfortable with backporting the movr workload impl into 19.1 (#37709 and #38203)? It clearly falls in the "new feature" bucket, but it's hard to imagine where any stability risk would come from, as it's just adding another implementation of the workload.Generator interface. Those PRs do touch a few files outside of pkg/workload/movr, but I could easily skip over those in the backport.

I think it's worth considering this PR separately, given that there's a very easy workaround (cockroach demo movr instead of cockroach demo) and changing the behavior of cockroach demo could potentially break someone's script.

[bdarnell]

I don't mind too much because it's essentially zero risk, but it does seem a little late to change this. It's too late for 19.1.4, so this will go into 19.1.5 in September. Is it worth the trouble to get this out for just a month before 19.2?

People don't upgrade often, but that also means that people will be sticking with 19.1.3 so we'll have to be careful in the docs to point out that this will require 19.1.5+. The target audience for this feature would mainly be new users, who will presumably be starting with 19.2 as soon as it's available.

I think it's worth considering this PR separately, given that there's a very easy workaround (cockroach demo movr instead of cockroach demo) and changing the behavior of cockroach demo could potentially break someone's script.

I'm not too concerned about scripting of cockroach demo, and I wouldn't want to introduce cockroach demo movr in 19.1.5 if it just changes back to cockroach demo in 19.2.

[jseldess]

Good points, @bdarnell. If it's unlikely for users on 19.1 to upgrade to 19.1.5, then this isn't worth it.

[awoods187]

@bdarnell what's your evidence that people won't upgrade to 19.1.5? I think many customers will not join 19.2.0 which means 19.2.1 isn't until November. It seems like this would have value in the meantime. In addition, plenty of customers will continue to use 19.1 even into 19.2.x.

[bdarnell]

Just anecdotal; feel free to pull actual data here. But even if people do upgrade quickly to 19.1.5, the marginal value of getting movr-based documentation out there a couple of months early seems fairly small to me.