Initial macOS ansible scripts

[gdams]

Okay so here are my initial ansible scripts for macOS. Obviously the inventory isn't complete yet but I will work with @mhdawson to get that side ready. I have confirmed that this works with a blank machine from macstadium (also ran a test build to confirm all tools were in place).

Things to note:

this requires NOPASSWD to be added to the sudoers file to enable elevation

sudo visudo
%admin          ALL = (ALL) NOPASSWD:ALL

Possible things to add?

do we want to disable the firewall on the test machines to prevent tonnes of popups?

[joaocgreis]

I don't think start.sh should be in all machines. There should be some kind of mechanism everywhere to restart Jenkins when it crashes, which is (or was) a relatively frequent issue. So, if Jenkins does not start because of the restart mechanism, we should debug the restart mechanism. If Jenkins starts and we want it stopped for some reason, we must always disable the restart mechanism anyway. Having start.sh everywhere will be very misleading and may not be kept in sync with the other way Jenkins is launched.

start.sh was traditionally used with monit. Here, I see it used with launchctl and a KeepAlive=true config entry. If that does not work, then we should debug it, not start Jenkins manually and risk having two running.

About this PR, shouldn't the two files under ansible/partials/ be in ansible/roles/package-upgrade/files/ instead, since they're only required for package-upgrade?

This shouldn't land with the current inventory. A partial one is ok, or even no inventory entries. Also, Gibson raised a valid point about the duplicated sudo entry. After these are addressed LGTM. Thanks for doing this!

[gibfahn]

I don't think start.sh should be in all machines. There should be some kind of mechanism everywhere to restart Jenkins when it crashes, which is (or was) a relatively frequent issue. So, if Jenkins does not start because of the restart mechanism, we should debug the restart mechanism. If Jenkins starts and we want it stopped for some reason, we must always disable the restart mechanism anyway. Having start.sh everywhere will be very misleading and may not be kept in sync with the other way Jenkins is launched.

I was thinking that maybe ~iojs/start.sh could just call the relevant service restart command (to avoid people having to remember what it is for each platform). Might be confusing to call it start.sh in that case though.

[joaocgreis]

I was thinking that maybe ~iojs/start.sh could just call the relevant service restart command

That would be quite nice! The jenkins-worker role should probably be separated by the boot system into partials, perhaps with a common section. It would be nice to include a start.sh (or restart?) script that does the right thing there, different for each start system.

[mhdawson]

Agreed it would be nice if there was a script that could be used to start/stop which calls the version for the specific platform. Maybe "agent-control" with start and stop options ?

[refack]

Can we land this, and iterate later? On just land the inventory?

[refack]

BTW the libuv jobs need GYP to be in ~/gyp so meanwhile I added a curl call in the job:
https://ci.nodejs.org/job/libuv-test-commit-osx/jobConfigHistory/showDiffFiles?timestamp1=2017-11-22_22-16-55&timestamp2=2017-12-01_15-30-59

[mhdawson]

I'm all for landing minus the inventory and then updating once we have additional changes (like adding in support for ccache)

[gibfahn]

I'm all for landing minus the inventory and then updating once we have additional changes (like adding in support for ccache)

SGTM, but it'd be good to add some TODOs to the source code (like for adding ccache) so we've noted the gaps.

[mhdawson]

It was just waiting on a fix up for cache. But I'm +1 for landing and then fixing that

[gdams]

So we are able to easily install ccache. My only remaining issue is adding ccache to the path of the jenkins node. I added the path to ccache in etc/paths and this works locally but I am unable to make this extra path appear in jenkins. (I have even tried restarting the machine)

[gibfahn]

So we are able to easily install ccache. My only remaining issue is adding ccache to the path of the jenkins node. I added the path to ccache in etc/paths and this works locally but I am unable to make this extra path appear in jenkins. (I have even tried restarting the machine)

@gdams is /etc/paths definitely added even when things are launched directly (not through a terminal)?

[gdams]

is /etc/paths definitely added even when things are launched directly (not through a terminal)?

@gibfahn I'm not 100% sure. I'll do some investigation.

[rvagg]

I needed the macstadium hosts in my ssh config and @Trott couldn't figure it out to go fix some broken hosts so I've pulled out the inventory.yml and plugin fixes here and pushed them to master as 0495bc2. It'd be good if we could get this merged soon, it can be iterated on once merged if it's not fully working yet.

[mhdawson]

I'm +1 on merging and then updating if we need to. @gdams can you rebase and then we'll do that?

[gibfahn]

It'd be good if we could get this merged soon, it can be iterated on once merged if it's not fully working yet.

Yep, we an just leave TODOs where things aren't working yet (or need manual setup).

[gdams]

updated PTAL

[gdams]

@mhdawson I think those merge issues are now cleared up

[mhdawson]

Ok, will try to land by the end of week unless there are objections, comments from other @nodejs/build team members.

[mhdawson]

@gdams can you squash your commits. git am is failing with conflicts even with -3. I suspect that if you squash that will get rid of the issue.

[gdams]

@mhdawson rebased and squashed

[mhdawson]

Landed as eb68f59