Correct way to use deploy and/or active project

Hi all,

I am not an experienced perl user, and I am doubly not an experienced ActiveState perl user…

I do have some experience deploying packages and languages to machines however.

My problem is: I cannot work out what the intended usage is for projects and deployments.

The history is:

  • I an resurrecting an old C++ build process that used to run using now obsolete OS/tools/SDKs etc
  • that used an earlier version of ActiveState perl and installed it by running an installer checked-in to the depot
  • I started looking into using a newer version of perl and got stuck as the way of installing packages suggested on the web-site seemed very cumbersome and awkward… not least because the wizard knew nothing of some of the packages I needed… (I subsequently found another bit of UI that did know about those packages but that wasn’t well advertised.)
  • after some head scratching I discovered “state install XXX” and that was good
  • and that prompted me to “push” and “activate” the project…
  • however newly installed packages frequently did not seem to be accessible to other processes on my system, it looked as if those processes were still accessing the default project (although running “state show” for them claimed they were not…)
  • THEN I discovered “state deploy” and that seemed to fix that, skipping whatever was going wrong with project configuration/caching and creating a folder with the installed packages explicitly copied into it…
  • (although my install cycle is now: state install XXX; state push; state deploy AAA/BBB – and this duplicated the folder entries in the path every time I ran it, requiring me to do manual clean-up, which is a pain and should be something the gets right itself… the concept of “deployment” usually comes with a assumption of idempotency so that it can be repeated to get incremental changes…)

So far, so good, my investigation moved on…

HOWEVER, now I have had cause to nuke that copy of the depot and make a fresh one (basically wanting to see how it behaves without some other changes I made) and I have encountered something I cannot quite get my head around…

  • when I type “perl” it tells me it “cannot find the project file” and it cites the (now nuked) location where the project was before I started “deploying” it elsewhere…
  • (I do not have that project’s folder anywhere in any paths anymore)
  • I wondered if maybe, now I was deploying, I did not need an activate project? But there does not seem to be any way to deactivate it…
  • I also hypothesized that I only needed the activated project to push from and that what I was doing with my install/push/deploy cycle was: 1) updating a project I was not actually running, 2) pushing that, 3) deploying to back to the location where it was run from – which I imagine should not be necessary but works.

So my question is How should I be using this. The situation is that I will configure this on my machine but, when others join the project, they will need to get the same configuration as I have. I was imagining we could do that by telling them to deploy the same project I was (or even just check-in the deployed folder, tying the versioning together in a very desirable manner). However, if they need an “activated” project as well as the “deployed” one I am not sure where that leaves us…

I have not, as yet, been able to switch the “activated” project to the same folder as the “deployed” one… and I am unsure whether this 1) makes no sense, or 2) would make sense but cannot get past the missing project folder (I can put that back, but I am still trying to understand what should be happening…)

I have looked for online help on this, but I’m finding things that say “this is how you activate” without explaining exactly what it is for, or how it interacts with deployment.

Any info/links/suggestions gratefully received.

Thanks,

Ian

Hi @IanBadcoe ! Thanks for the very detailed explanation – you’re actually totally right in that your explanation:

The situation is that I will configure this on my machine but, when others join the project, they will need to get the same configuration as I have. I was imagining we could do that by telling them to deploy the same project I was

Is totally correct – there’s just a few details I think that are dependent on your exact configuration. So in general, to achieve the above, what you would do is:

  • Create a project – normally as you found out, you start from the Web UI. You can then add the packages you want either via the web UI or via state install. The reason you want to create a project is that it creates a shared source of truth for what your configuration is amongst your team members. It’s also revisioned so that it’s always reproducible.

  • Anyone who wants to use that configuration simply activates the project, by typing state activate myOrg/myProject – this will install everything you set up in your project on their machine. Now, here’s where I think there might be confusion – unlike “installing” Perl on your machine via an installer, what activate does, is it sets up a virtual environment–like a sandbox that is a totally self-contained Perl installation. This means that if you want to switch projects to another Perl with a different set of packages, you can easily do so just by “activating” a new project. It’s also worth nothing that the project is only “activated” within the shell that you’re currently running, so if you have other processes that need to access Perl, that’s where deploy comes in.

  • If you need a sort of classic “installation” of Perl that can be globally accessed on your machine by other processes, etc. then you can use state deploy and this will basically “install” Perl on your machine. You might need to setup some PATH variables to point it at the installation like PERL5LIB, and add the location where you deployed it to your PATH so that it can find perl.exe, etc. But once deployed you do NOT need to activate.

  • The other thing you mention is state push – so, if you’re familiar with git – this is the same concept. When you run state install – any changes you’re making are ONLY on your local machine. If you want to “push” them to the platform (at which point they’re shared with all your team members), then you run state push and it will push those to the platform.

Hope that helps clarify some of the concepts used on the platform – happy to explain more if there are other questions.

1 Like

Hi Pete,

Thanks you, that is clear, however I think there are some other complexities going on.

A thing I forgot to mention in my original explanation is I also tried “state activate IanBadcoe/perl --default” which looked as if it should make the project universal across the whole machine? After doing that things looked confused and other processes (even after a reboot) reported being set to my project but could not find the recently installed modules… when I looked at the @INC value reported in the errors, it was using a cache folder dated back when I first installed Perl, instead of the newer one which (I presume) went with the current project.

Because in this case there are a large number of scripts/vc-projects/configuration files etc etc, any one of which might invoke perl, I cannot practically add a “state activate” to every place where it might happen, so from that point of view I think deploy is the way to go (unless “activate --default” effectively has the same effect – but, as I said, it seemed inconsistent when I tried that…)

I am also not sure how to handle things if I have an activated and a deployed project on my system (is this a bad state to be in, or an expected use case?). Because of the history I followed (with “activate --default” AND “deploy”) I think this is the state I find myself in. I had been assuming that the deployed copy completely took over from the activated one, but the current error about not being able to find the project folder implies otherwise… (and I have not found how to remove a default activation.)

Thanks again, this is clearer, and I think deploy is the way to go for other users, but I am still a little uncertain how to resolve the situation I have on my machine where I am running installs/pushes as well as consuming the deployment…

Cheers,

Ian

Thanks for the additional info @IanBadcoe – it sounds like there might be a bug with the --default parameter as it should behave the way you expect where it’s effectively the same as a deploy, but with the added benefit that it’s easier to keep updated compared to the deployment (which you’d have to re-deploy each time you made a change to the master project). Let me file an issue for the --default problem with the dev team here as your INC folder should definitely be updated.

1 Like

Thank you!

I think it all makes sense now…

I was seeing what I thought was another bug, where when I redeployed it retained some obsolete paths from the previous deployment in PATH, but I have realised that was because the folder where I was deploying was not empty, and had previously been configured from a different location (there are some symbolic links involved in the depot configuration). Once I wiped the folder, the paths came up correctly… so I think that is odd behaviour (since the deployment knows where it now is and could presumably therefore get it right) but arguably encountering this is user error rather than a bug.

I think this sorts me out, thanks! I will have to continue the install->push->deploy cycle, which is not quite as convenient as “activate --default”, but I can get by.

Thanks again,

Ian