Alertmanager: Deprecate (but keep compatibility) of cluster flags

[gotjosh]

What this PR does:

Deprecates flags with the cluster prefix within the Alertmanager. I've
opted for a "high-availability" nomenclature which feel a bit more
aligned with what the flags are actually doing.

Signed-off-by: gotjosh josue@grafana.com

Which issue(s) this PR fixes:
Fixes #3670

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[gotjosh]

I spent some time trying to debug the integration test w/o much luck. It seems like the DNS resolver for docker is not working as expected on CI as the past pass locally. My understanding is that Docker uses the host's /etc/resolv.conf but I'm not able to inspect that file by logging on the machine directly - any help?

[pracucci]

You did great! LGTM with a couple of nits.

I'm now going to look into the integration test issue.

[pracucci]

I'm now going to look into the integration test issue.

To keep compatibility across platforms, a container needs to be address by the host <network name>-<container name>, which is what the e2e framework does when you use functions like NetworkEndpoint(). However, such functions can't be used in this specific integration test because you have chicken-egg problem: to call NetworkEndpoint() you need the service first, but you need to create the first you need you need the peers list first.

I've refactored the code and fixed it (see commit).

[gotjosh]

which is what the e2e framework does when you use functions like NetworkEndpoint()

Thanks @pracucci, I initially suspected it had something to do with that but when I inspected the return value of NetworkEndpoint() locally in the debugger it did not have the network name as a prefix so I quickly discarded it. 🤦

[pracucci]

LGTM, thanks!

[pstibrany]

LGTM. Is it worth mentioning that old flags still take precedence, if set to non-default values?