Hi
Thank you very much for the reply and the interest in helping.
vtamara: did previous versions of Ruby, unicorn, or OpenBSD work
for your app?
The application was working withouth issue on OpenBSD/adJ amd64 6.7,
ruby 2.7 and unicorn 6.0. So there is something in conflict betweeen
Unicorn 6.0 with OpenBSD/adJ 6.8 and ruby 3.0.
Is preload_app in use? If so, does the problem manifest without
preload_app?
No
Are you using Ractor or any other new/experimental stuff
in Ruby?
No
Can you reproduce the problem with a simple "hello world"-style app?
I don't have a hello world example at hand, but some projects that
present this problem when I run them on OpenBSD/adJ amd64 6.8, nginx
1.18.0, node 12.16.1, PostgreSQL 13.2, Ruby 3.0, Rails 6.1 and Unicorn
6.0 are:
https://github.com/pasosdeJesus/cor1440
https://github.com/pasosdeJesus/cor1440_pdJ
https://github.com/pasosdeJesus/sivel2
https://github.com/pasosdeJesus/si_anzorc
https://github.com/pasosdeJesus/si_asom
https://github.com/pasosdeJesus/si_codacop
https://github.com/pasosdeJesus/sivel2_mujeresindigenas/
The simplest of them is https://github.com/pasosdeJesus/cor1440, althoug
it requires:
* PostgreSQL
* nginx
* nodejs
* yarn. I install it with: doas pkg_add bash; ftp -o-
https://yarnpkg.com/install.sh | bash
* A directory /var/www/bundler/ruby/3.0 where I install the gems with
the permissions of a user (not root) that runs the application, and
better running:
bundle config path /var/www/bundler/.
* unicorn installed globally with:
doas gem install unicorn
doas ln -s /usr/local/bin/unicorn_rails30 /usr/local/bin/unicorn
To create the PostgreSQL user (in the following example sipdes) and a
database for the production environment (in the following example
cor1440gen_pro):
doas su - _postgresql
createuser -h /var/www/var/run/postgresql/ -U postgres -s sipdes
createdb -h /var/www/var/run/postgresql/ -U postgres -O sipdes
cor1440gen_pro
psql -h /var/www/var/run/postgresql/ -U postgres
postgres=# alter user sipdes with password 'xyz';
postgres=# \q
exit
With nginx suppossng I will run the application in localhost over SSL,
what I do is to choose an unused port (let's say 2009) and add to the
/etc/nginx/nginx.conf configuration in its http block:
upstream unicorncor1440 {
server 127.0.0.1:2009 fail_timeout=0;
}
I also add a server block inside the http block with:
server {
listen 443 ssl;
ssl_certificate /etc/ssl/server.crt;
ssl_certificate_key /etc/ssl/private/server.key;
root /var/www/htdocs/cor1440/;
server_name 127.0.0.1;
error_log logs/cor1440.rror.log;
location /cor1440 {
try_files $uri @unicorncor1440;
}
location @unicorncor1440 {
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_set_header Host $http_host;
proxy_redirect off;
proxy_pass http://unicorncor1440;
error_page 500 502 503 504 /500.html;
keepalive_timeout 10;
}
location ^~ /cor1440/assets/ {
gzip_static on;
expires max;
add_header Cache-Control public;
root /var/www/htdocs/cor1440/public/;
}
location ^~ /cor1440/images/ {
gzip_static on;
expires max;
add_header Cache-Control public;
root /var/www/htdocs/cor1440/public/;
}
location ^~ /cor1440/packs/ {
gzip_static on;
add_header Cache-Control public;
root /var/www/htdocs/cor1440/public/;
}
}
And restarted nginx:
doas rcctl restart nginx
Then to configure:
cd /tmp
git clone [email protected]:pasosdeJesus/cor1440
doas mv cor1440 /var/www/htdocs/
doas chown -R $USER:$USER /var/www/htdocs/cor1440
cd /var/www/htdocs/cor1440
mkdir .bundle
echo '---' > .bundle/config
echo 'BUNDLE_PATH: "/var/www/bundler"' >> .bundle/config
echo 'BUNDLE_DISABLE_SHARED_GEMS: "true"' >> .bundle/config
bundle
Here some gems will require installing with special permissions for
example:
doas gem install --install-dir /var/www/bundler/ruby/3.0 unicorn
Other gems that will require this kind of installation will be:
racc nio4r websocket-driver bcrypt bindex msgpack ffi redcarpet kgio
libxml-ruby pg
nokogiri bootsnap puma sassc
Once bundle runs completely:
cp .env.plantilla .env
vi .env
There I set the variables:
* BD_USUARIO with the database user,
* BD_CLAVE with the password
* BD_PRO with the name of the production database
* CONFIG_HOSTS=127.0.0.1
* DIRAP with the path to the application
* RAILS_ENV=production
* PUERTOUNICORN with the port I choosed and set in nginx.conf
* USUARIO_AP with the operating system user who will run the application
Prepare credentials and database:
bin/rails credentials:edit
RAILS_ENV=production bin/rails credentials:edit
bin/railsp db:setup db:migrate sip:indices
Precompile assets:
CXX=c++ yarn
doas ln -s /usr/local/bin/node /tmp # To overcome bug of node with
rails on OpenBSD
bin/railsp assets:precompile
Run the application with:
SECRET_KEY_BASE=cf46 DIRAP=/var/www/htdocs/cor1440/ bin/u.sh
Trying to see the application running with a browser in
https://127.0.0.1/cor1440 will produce the errors described in
log/unicorn.log.
Blessings. Thank you.
El 2021-03-30 00:00, Jeremy Evans escribió:
On 03/30 01:58, Eric Wong wrote:
vtamara <[email protected]> wrote:
> Hi
>
> Launching a rails application with Ruby 3, Unicorn 6 on OpenBSD/adJ 6.8, I'm
> experiencing a lot of "Bus Error" and "Segmentation Faults" during
> aproximately half hour.
+Cc: Jeremy Evans, who is more familiar with OpenBSD and Ruby 3.0 than
I
cf.
https://yhbt.net/unicorn-public/[email protected]/
I'll be happy to look into this if a minimal self-contained
reproducible
example is posted. Personally, I run my Ruby apps on OpenBSD amd64
with
Unicorn, and haven't seen similar errors. However, I don't use
preload_app and use fairly minimal dependencies.
Thanks,
Jeremy
--
Dios, gracias por tu amor infinito.
--
Vladimir Támara Patiño. http://vtamara.pasosdeJesus.org/
http://www.pasosdejesus.org/dominio_publico_colombia.html
--
unsubscribe: one-click, see List-Unsubscribe header
archive: https://yhbt.net/unicorn-public/
