increase default instance root volume size

[chrislovecnm]

Fixing #1929

masters will default 50 GB
nodes will default to 100 GB

50 and 100 just seemed weird ... doing 64 and 128 next rebase ...

---

This change is 

[geojaz]

sorely needed. good stuff lgtm, just fix the tests for volume size :)

[justinsb]

I was thinking "we can't change the defaults", but I guess we can actually in a major release if we think it is a better default, and we have a release note that calls it out clearly.

i.e. not specifying means "omakase" and we can change things as long as the change is clearly signposted and unlikely to actively cause problems.

In this case, the only damage is finanical. It is $10 per month instead of $2 per month, per instance, so we should explain it in the release notes, but if you & geojaz both think it a better default, it works for me.

[justinsb]

Oh, and the three benefits from a bigger disk:

• more disk space (duh!)
• more inodes (because of the filesystem)
• more iops (because of the gp2 instance type)

A potential downside is that a bigger disk might mean more scanning, and we've seen reports of that getting slow, but I think we need to address that anyway (in k8s/k8s) if it turns out to be an issue.

[justinsb]

lgtm once tests passing. If you want to start a "release-notes-1.7" doc alongside the 1.6 one as well that would be great :-)

[chrislovecnm]

The one time I don't run ci - heh

[chrislovecnm]

@justinsb / @geojaz - I increase the size to 128 nodes, 64 master, 32 bastion. Having odd HD sizes like 50 and 100 was hurting my head. CI is passing.

[justinsb]

One doc change needed then lgtm (i.e. self merge after doc change)

[justinsb]

64/128

[justinsb]

Merging, will send doc fix.

[justinsb]

Doc fix in #2873