GemStone on ubuntu server
Locked
 
|
Hi guys,
I wonder how should I install GemStone on ubuntu. Let me give you context so I can better ask you for your advice: We are developing the beta version of a SaaS and we have a setup that has 3 parts: 1. local develop environment (macOS) 2. staging server (publishes automated updates when pushing to the develop branch) 3. production server (this is a different physical host than staging and publishes automated updates when pushing to master branch) I’m thinking in doing an installation of gemstone in the staging host and another in the production host. Both have ubuntu 16.04 and I think the best is each to have 1 stone of its own. So far I’m good with my local dev setup, so now I wonder a bit on how I’d do the devops for the GemStone part. So questions: 1. What’s the recommended way to install a small stone with seaside REST and NeoJSON for those hosts 2. How can I use MockGemstone on Pharo, save the packages and get them into the stone? can be automated? 3. Do you have any suggestion to make a CI to trigger an update on the server when I push a commit to develop or master? (I can make it at the beginning no problem but I’d try to be automated instead of human intervention on those updates) Thanks! _______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
Re: GemStone on ubuntu server
|
|
On Mon, Jan 9, 2017 at 2:03 PM, Sebastian Sastre via Glass <[hidden email]> wrote:
Probably you want to use gsDevKit_home for creating the stones [1] and likely those for seaside [2].
What I do, is to have all those packages (including MockGemstone) in my Metacello conf ConfigurationOfMyApp. So... all I have to do on the server is a metacello load.
Not 100% to answer your question, but you may want to take a look to this CI project as it works in Pharo and GemStone [3] Cheers,
_______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
Re: GemStone on ubuntu server
|
|
On Mon, Jan 9, 2017 at 2:21 PM, Sebastian Sastre <[hidden email]> wrote:
Yes. Automated? half and half. I do have quite some bash scripts to do this, but as it happens every in a while and I have lots of "none default" gemstone / app details, it happens to me that each time I need a new stone installation, some part of the scripts fails and I need to update some scripts. Because of this, it's human in the sense that I manually invoke each part (a bash script) of the process...
_______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
Re: GemStone on ubuntu server
|
|
In reply to this post by GLASS mailing list
Hello Sebastian, You've asked some fairly broad questions here and I'm hoping that other developers will chime in with some details about the approaches that they've taken (those that have worked and those that haven't). For the production and staging servers, you will probably need to
some sort of process monitoring. On http://ss3.gemtalksystems.com/, which is running Ubuntu 12.04,
I use daemontools[1] for automatic restart of gems, monit[2] for
ensuring that the running seaside servers are actually responding,
and custom scripts in /etc/init.d for starting the system when the
server reboots. The ss3 installation predates my GsDevKit_home work, but today I would base the ss3 installation on GsDevKit_home. I know that Paul DeBruicker has been doing work with integrating daemontools into the GsDevKit_home environment[4] and building some ansible[6] scripts that are aimed at deplying GsDevKit_home environments[5] ... I know that these are currently a work in progress, but perhaps Paul can provide some additional pointers? ======= In the following section, I want to dive a little deeper into
some of the GsDevKit_home features that may apply to the
management of your dev/stagin/production servers. If you look at at GsDevKit_home installation, you will note that the directory structure under $GS_HOME/sys/local is a separate github project GsDevKit_sys_local. $GS_HOME/sys/local is where you store your local GsDevKit_home customization. For your system, you can add your personal git projects so that they will be managed consistently by both the $GS_HOME/bin/updateGsDevKit and the tODE `project list`. The $GS_HOME/sys/default/gsdevkit_bin repository management scripts[10] (called by $Gs_HOME/bin/updateGsDevKit) can be customized by putting local versions of the scripts into the $GS_HOME/sys/local/gsdevkit_bin directory. The tODE build scripts[11] can be customized by putting customized copies of the scripts into $GS_HOME/sys/local/client, or adding your own client scripts. For an example of tying into the $GS_HOME/bin/updateGsDevKit, you can look at how I customize the GsDevKit_home travis test environment[7]. The tODE build scripts[8] are customized to remove the backup step and the update scripts[9] are customized to control which branch of the shared git projects should be used. The $GS_HOME/sys/local/server/home directory is where shared tODE scripts can be stored ... a tODE script stored here will be seen in the /home node in all stones. The $GS_HOME/sys/local/server/projects directory is where shared project entries are stored ... a project entry store here will be seen the `project list` and /projects node for all stones .... Since the $GS_HOME/sys/local is a separate git project it is easy for you to manage your git clone and share it between your dev/staging/production machines... the project entries and gsdevkit_bin scripts are likely to differ between each of the systems, but that can be managed by keeping three or 4 different branches: one branch for each of the systems scripts ... depending upon how you want to manage things, a 4th branch can be used for the common entities ... ======== For building customized stones, I would point to the Tugrik[18] and Gt4Gemstone[19] as good examples. In both of these cases an URL is used to access the project entry for the project, but in you case you can reference a local project entry. See the gh-pages branch of GsDevKit_home[20] for additional project entry examples. Once you've created a custom stone, you can use $GS_HOME/bin/todeIt to trigger stone update by running the `project load` tODE command: $GS_HOME/bin/todeIt <stone-name> project load <project-name> If you want to run a tODE script from bash, then you can do the following: $GS_HOME/bin/private/gsDevKitTodeCommandLine todeIt <stone-name> << EOF project load <project-name-1> projet load <project-name-2> # additional tODE commands EOF If you prefer/need to run Smalltalk from bash then the following can be used: $GS_HOME/bin/startTopaz <stone-name> -l << EOF login run "arbitrary smalltalk" % commit exit 0 EOF ======== With regards to triggering automatic updates, there is nothing out of the box for doing this... GitHub has the WebHooks service[12] where an url is hit with an http post[13] allowing the server to take whatever action is needed ... I haven't looked into this in detail, but I've always figured that it would be interesting to eventually hook into this service. There are a couple of Pharo projects[14] and [15] that implement the GitHub API[16] as well. The WebHooks server could be implemented in either Pharo or GemStone. GsDevKit_home uses Pharo already for implementing some of the GsDevKit_home scripts[17], so Pharo "is aware" of the GsDevKit_home directory structure and scripts. ======== Hope this helps you get started, Dale [1] https://github.com/Monty/GemStone_daemontools_setup [2] https://www.mmonit.com/monit/ [3] https://github.com/GsDevKit/GsDevKit_sys_local [4] https://github.com/GsDevKit/GsDevKit_daemontools [5] https://github.com/GsDevKit/GsDevKit_ansible [7] https://github.com/GsDevKit/GsDevKit_home/tree/master/tests/sys/local [8] https://github.com/GsDevKit/GsDevKit_home/tree/master/tests/sys/local/client/tode-scripts [9] https://github.com/GsDevKit/GsDevKit_home/tree/master/tests/sys/local/gsdevkit_bin [10] https://github.com/GsDevKit/GsDevKit_home/tree/master/sys/default/gsdevkit_bin [11] https://github.com/GsDevKit/GsDevKit_home/tree/master/sys/default/client [12] https://developer.github.com/webhooks/ [13] https://developer.github.com/webhooks/#wildcard-event [14] https://github.com/astares/GitHub-API [15] https://github.com/Balletie/GitHub [16] https://developer.github.com/v3/ [17] https://github.com/GsDevKit/GsDevKit_home/tree/master/shared/repos/todeClient [18] https://github.com/dalehenrich/Tugrik#install-gsdevkit_home [19] https://github.com/feenkcom/gt4gemstone#installation [20]https://github.com/GsDevKit/GsDevKit_home/tree/gh-pages On 1/9/17 9:03 AM, Sebastian Sastre via
Glass wrote:
Hi guys, _______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
Re: GemStone on ubuntu server
|
|
In reply to this post by GLASS mailing list
Hi Sebastian.
I have some ansible (https://www.ansible.com) scripts that should do what you want as far as taking a blank Ubuntu server and getting GemStone running under daemontools and behind nginx. You can load a project from a Baseline into the stone or copy up a stone from your dev environment. I've been waiting to publish them because I haven't made time to get the hot standby working yet. Let me double check I don't have any sensitive info in the repo and I'll publish them sometime in the next couple days. I write stuff mostly in Pharo and only use MockGemstone in the Pharo images. It just provides stub classes for GemStone functionality so you don't need it in the GemStone stone. Be sure to run all your file system and socket access through it in Pharo. As for updating your stone after a git commit you ought to be able to specify a post-commit hook on your laptop and post-receive hook on the server to get the stone to update to new code without involving a CI unless you want to. This is a good guide http://toroid.org/git-website-howto for getting a git repo populated with updates after a git commit. In that post-receive hook they desrcibe you could have curl hit a URL the stone is watching to trigger the Metacello or Gofer load of your new code. The post-commit hook on your laptop pushes the code to the server. The post-receive hook populates the git repo on the server that the stone uses to look for code and then the curl GET triggers the load. One problem I haven't yet figured out is how to roll back to an earlier SHA when I commit bugs. I just do it manually, but would be happy to learn another way. Hope this helps Paul
|
Re: GemStone on ubuntu server
|
|
Hey Paul!
Your workflow sounds really good! Couple of ideas: 1. have in mind that ubuntu 16.04 (TLD) switched to systemctl for managing services. That change was a PITA, I had to change scripts for many of my other services. Anyway, I don't see any blocker with GemStone, with a couple of env vars and the executable line all should be up/down easily 2. the ansible script to build a server should be super nice to have. Did you publish it somewhere? 3. For the commit/hook, I see what you mean but I think is easier to use CircleCI (or any able CI) to react to the commit, actually no need for a build phase, and go directly to do the deploy running a script that loads the new mcz or filetree package in the stone. Heh, it sounds easier than many javascript projects I deal with daily :) About pushing bugs and rollback, it comes to my mind that one could have a filetree git clone somewhere in the server and a script to change gears to a given git commit SHA and suck it into the stone. Would still be manual, so not sure if that might save you of anything. Thanks for sharing your workflow, looking forward to hear about how you use ansible for this Seb |
| Free forum by Nabble | Edit this page |