Standardise hostnames on our machines

[sxa]

Follow-on to #131 - we should define a standard machine naming convention. Currently most are of this format:

<usage>-<provider>-<arch>-<os>-<sequence>

Although some are missing the architecture - mostly windows/macos boxes where they're all x64. We do also have an issue where some of our tests don't work properly when there is a . in the machine name (for example those with names includeding ubuntu-16.04) and in those cases we have replaced the . with a - on the machine itself.

This work item will discuss that convention and it is what's used in jenkins, nagios etc and listed on https://github.com/AdoptOpenJDK/openjdk-infrastructure/blob/master/infrastructure.md, and whether the hostnames on the machines we get from the providers should be changed to match these (possibly with the . to - translation ... Or whether we should just avoid having . in the names at all. At the moment some of the hostnames on the machines don't match what we call them, which is inconvenient when you're logged into multiple machines

[bblondin]

yes I agree the . periods do not translate. using - dashes is better
Issue #165 (Set system hostname to match jenkins) is related to this

[judovana]

I would wote for . replaced by _ instead by -.
reason is ubuntu-16.04 would change to ubuntu-16-04 where the . had different meaning then have new dash.

Generality strict delimiter of various parts is currently dash. So it should not be used in sub parts.

Also, what is and .. well and ... for?Well For us newbies, can you elaborate more on individual parts of ----?
Shouldn't be some of those handled elsewhere? Ideally all except - ?

[sxa]

@judovana I'm not sure what you mean by the "what is and ... well and ... for?" question. Or "individual parts of ----?"

[judovana]

Ha. Markdown had eaten the tags. Sorry
Also, what is <usage> and <prvoder> .. well and <sequence> ... for?Well For us newbies, can you elaborate more on individual parts of <usage>-<provider>-<arch>-<os>-<sequence>?
Shouldn't be some of those handled elsewhere? From those probably only <arch>-<os> are celar?

[sxa]

Ah :-D <usage> is mostly one of build test jck, <provider> is whichever cloud provider the machine comes from so softlayer marist joyent etc. <sequence> is just a number to distinguish them if we have more than one of the same machine so -1 -2 etc. The link in the initial post has all the machines so has various examples.

[judovana]

Thank you. My nit would be that usage should be handled differently. Machine should (== one wishes) transparent. Provider is interesting, and sequence made sense. TY!

[gdams]

closing this as we have now got our inventory.yml sorting this out. I will create new issues to bring any old machines up to date with the new hostname schema