Hi Tim

Here are my notes on installing Pharo in a DigitalOcean Ubuntu droplet.
I usually go thru a Windows Command prompt box having installed xfec4 in
the ubuntu droplet, but the command line connect and graphical remote may
be easier for a Linux users. my ssh public key is also in the DO droplet
Now the first step for me is a installing Pharo launcher thru command line
and then everything is thru graphical interface

Installing and checking Pharo-Launcher, Installing Pharo 8 64 bit from
pharo.org (instructions as on Pharo.org)

1. In Windows 10 command prompt connect thru > ssh root@<Droplet-ip>
2. cd
3. curl -o pharo-launcher.zip -L
   https://files.pharo.org/pharo-launcher/linux64
4. unzip pharo-launcher.zip
   or thru the GUI-> extract here
   (pharo-launcher files will be extracted in ./pharo-launcher)
   Now while connected to the linux graphical interface thru windows remote
   terminal and in the GUI
5. Create a icon on desktop thru right-click “Create Launcher” for
   pharo-launcher
6. Create pharo images thru pharo-launcher

hope this is of use

Sanjay Minni

On Thu, 1 Apr 2021 at 16:31, Tim Mackinnon tim@testit.works wrote:

Thanks Sanjay - you have reminded me that I have some similar notes somewhere (now located) that did the command line foo to get things running - looking at mine there was quite a big of dance to provide a way to gracefully stop and start the image so that you can easily and automatically redeploy your changes (read: use Github  actions or Gitlab CI).

So I’m curious on whether Docker is now sufficiently stable stable/easy/cheap to make it a viable alternative - and whether that is also cost efficient.

Pablo wrote a recent blog post on running Pharo in Docker using the BA images - https://thepharo.dev/2021/02/24/running-pharo-9-in-docker/ https://thepharo.dev/2021/02/24/running-pharo-9-in-docker/ - but while easy on the surface, if anything goes wrong - there seems to be very little debug output to know what has happened (I’ll post separately on this - as I’m looking at comparing options here).

With Docker options, I notice that dockerize.io http://dockerize.io/ (not used, just a quick search) - has a micro plan for $2/m - but is 500mb ram enough (there is a $5 one for 1gm ram).

Or - I stick with DigitalOcean and roll my own like before - and perhaps that has got a bit simpler.

I’m still curious what the wider community is doing.

Tim

I've been running a Pharo application since August last year,
initially setup with everything inside a Docker swarm (database,
reverse proxy and Pharo workers for the UI and REST API).

I wanted to give Docker a try because I thought it was going to be
"better". After having some issues with the Swarm networking affecting
PostgreSQL connections [1], I ended up having a hybrid that is not the
best of both worlds (I wouldn't say it's the worst either).

So my summary is that unless you have several host machines, having a
"swarm" is completely overkill and brings more friction than anything
else.
The plus of docker is that from Gitlab everytime I push a commit to
master, a new docker image is created and I can update things smoothly
on the server by pulling the image and upgrading the stack service.

"Whenever I have time" I plan to make the reverse proxy external to
docker (I currently use traefik as a container), and just keep the
database and Pharo workers as docker containers.

As for the stability, I only shut it down for upgrades, but in
February I tried with a single VM+image serving everything (just to
measure real use) and albeit it was a little slower, it handled
everything perfectly [2] without a single hiccup.

Regards,

[1] https://twitter.com/emaringolo/status/1296635983358763010
[2] https://twitter.com/emaringolo/status/1360247046553362432

Esteban A. Maringolo

On Fri, Apr 2, 2021 at 6:33 AM Tim Mackinnon tim@testit.works wrote:

esteban

I would like to have a success story for the Pharo web site
Could you come up with two paragraphs?

And if you have a visual it would be perfect.

Remember a success story is that you could do something not that your app become facebook.
S.

Stéphane Ducasse
http://stephane.ducasse.free.fr / http://www.pharo.org
03 59 35 87 52
Assistant: Aurore Dalle
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

Adding some thoughts to my own post - I have experimented with dockerize.io and so far the $2 plan (which so far I haven't been charged for) has been quite adequate for hosting something simple (and I suspect the $5 plan would do the trick as well).

It does get a bit more pricey when you add a database and unlimited traffic (60GB MySQL Database: 4GB, 1000m + no traffic + 60GB persistance + 100 mb logs = $26/m) - but even then thats pretty good.

It was relatively easy to do (doesn't have native Github integration yet, but has a cli for easy deployment) , and given what Esteban mentioned of his experience - this seems worth pursuing for ease of deployment/maintenance.

I did find Pablo's instructions were a bit limited - in particular covering a private repo (and how Pharo and the BA docker image give no output on credential failure - I've raised an issue on BA github, but may be a pharo thing - not yet sure - but it is quite easy to overcome). It also doesn't cover safe credential storage and best practices etc. (so hopefully a future article from someone more experienced in this?)

I'm curious if others have experienced anything similar they can share? I am a bit nervous the dockerize.io has little history (been around since 2019 allegedly) - but for my experimentation its fine. Others may need more, and I'm wondering where to help document this as deployment is a big stumbling block - that we actually can handle quite well with recent tooling.

In this vein - I wonder where SmalltalkCI sits? Do you need it for a plain Pharo app - you just need to grep the output of Metacello and your test runner  - but perhaps it does this better (I can see the need for building multiple platforms etc).

Tim

On Fri, 2 Apr 2021, at 10:32 AM, Tim Mackinnon wrote:

Thought it might be useful for anyone who has read Pablo’s blog post - this is a comment that I have tried to add (hopefully it will come through there):

Nice article which was very helpful in getting me quickly up to speed on deploying a Pharo app with Docker.  It might be worth adding a small addendum - when build doing a "docker build" I had some issues, and adding an additional switch "--progress=plain" gives much more verbose output that can help better see whats going on.

I also feel its quite important to call out that the "eval 'Metacello new..." command referenced in the DockerFile RUN - is only going to work on a public repository. This had me stumped when I tried the example on my repository - and its made particularly mysterious because the Docker output (even with --progress=plain) shows nothing - it just echo's the VM info text and then stops and hangs. I think this is something that should be addressed either by the BA Docker image (I've reported it as an issue there for now) or by Pharo (in my case it was P8). I am guessing that there is a credentials error/system prompter sitting waiting for user input. I think there should be a way to disable this behaviour somehow - but at the very least it should write something out to stderr so that you know something has happened?

Luckily there is an easy fix (for anyone wanting to load their own private project) - rather than using the "eval" command, use the "st" command and specify a load.st script. In this script specify some IceCredentialStore details eg:

IceCredentialStore current
storeCredential: (IceTokenCredentials new
username: 'email@mail.com';
token: 'xxxxxxxxxxxxxxx';
yourself)
forHostname: 'gitlab.com'.

Metacello new
baseline: 'myapp';
repository: 'gitlab://me/repo'';
onConflict: [ :ex | ex useLoaded ];
onUpgradeUseLoaded;
onWarningLog;
load.

Hopefully this might help anyone else directed here.

I also added some extra logging to my load.st - eg.

| logger |
logger := Stdio stderr.

logger crlf; nextPutAll: 'Starting Load Script...'.

Tim

On Wed, Apr 07, 2021 at 02:21:14AM +0100, Tim Mackinnon wrote:

Hi,

I build my Pharo image outside of Docker. My Dockerfile then packages the
built image plus whatever artefacts required into a Docker image. I am
using my Alpine Linux-based VM.

GH: https://github.com/pharo-contributions/Docker-Alpine/
DH: https://hub.docker.com/r/pierceng/pharovm-alpine

Pierce

Head slap.... I completely forgot I could do that, thanks for reminding Pierce (its almost better as well, as you can build your image, save it, then run tests on it, and then if it passes all your stages you can included the saved version in your docker build...  which sits better with me as it also means that log artifacts etc can be more easily examined for failures (and your not trying to console into a docker image to figure it out). As a nice side effect - if you are doing this in something like Gitlab - you have your project source already there and so you can avoid an extra checkout step, so its a little bit faster.

Pierce - who/where are you rnning your Docker image? Do you use a particular service (or are you running your own?)

Tim

On Thu, 8 Apr 2021, at 4:29 AM, Pierce Ng wrote:

On Thu, Apr 08, 2021 at 11:58:10AM +0100, Tim Mackinnon wrote:

I am using a Linux VPS which I manage myself.

The Pharo application is my blog. It was a web app fronted by a HTTPS
reverse proxy. At the start of this year I rewrote it into a static site
generator. Instead of running all the time, Pharo is now invoked when
triggered by a CI hook, which happens when I publish a new blog post or
otherwise update my site.

I play with all kinds of software on my VPS and it was getting a bit RAM
cramped, hence the rewrite. Wasn't willing to spend more money every month
for more RAM. :-P

Pierce