# py-push-server

## Docker Build & Deploy

- Update CHANGELOG.md

- Commit & tag with release version

```
export PUSH_SERVER_VERSION=0.1

# This command is also in build.sh
# Note that there is a user with the uid 1000 inside the Dockerfile: make sure that such a user exists on the server and has write access to the database.
docker build --tag py-push-server:amd-$PUSH_SERVER_VERSION --platform linux/amd64 .

docker save -o ~/dl/py-push-server-amd-$PUSH_SERVER_VERSION.tar py-push-server:amd-$PUSH_SERVER_VERSION

bzip2 ~/dl/py-push-server-amd-$PUSH_SERVER_VERSION.tar
```

... and then on the server after transferring that file:

```
export PUSH_SERVER_VERSION=0.1

bzip2 -d py-push-server-amd-$PUSH_SERVER_VERSION.tar.bz2

sudo docker load -i py-push-server-amd-$PUSH_SERVER_VERSION.tar

# start with an empty DB
cp data/webpush.db.empty py-push-server-db/webpush.db

# here's how to run it -- but if you're on production then look below about ADMIN_PASSWORD
sudo docker run -d -p 8900:3000 -v ~/py-push-server-db:/app/instance/data --name py-push-server-$PUSH_SERVER_VERSION py-push-server:amd-$PUSH_SERVER_VERSION
```

On a production server for security (eg /web-push/generate_vapid): set an environment variable `ADMIN_PASSWORD` for permissions; one way is to add this to the `docker run` command: `-e ADMIN_PASSWORD=<anything secure>`

Finally, after it's started, generate a new VAPID by hitting the `regenerate-vapid` endpoint with a POST, eg. `curl -X POST localhost:8080/web-push/regenerate-vapid`



## Docker Compose & HAProxy Setup

On a production server for security (eg /web-push/generate_vapid): set an environment variable `ADMIN_PASSWORD` for permissions; one way is to create a .env file with the value inside before running `docker compose` commands:

```
ADMIN_PASSWORD=<anything secure>
```

On first run you need to:

`docker network create phoenix-network`


### HAProxy setup

... in docker-compose.yml ...

```
  default-backend:
    container_name: 'default-backend'
    image: nginx:1.22.0-alpine
    volumes:
      - /docker-volumes/haproxy-config/core/nginx/html:/usr/share/nginx/html
    restart: always
    networks:
      - phoenix-network

  rsyslog:
    container_name: 'rsyslog'
    hostname: 'rsyslog'
    image: alpine-rsyslog
    build:
      context: ./alpine-rsyslog
    volumes:
      - $PWD/haproxy-config/core/haproxy.conf:/etc/rsyslog.d/haproxy.conf
      - $PWD/haproxy-config/log:/var/log
      - $PWD/haproxy-config/spool:/var/spool
      - $PWD/rsyslog/rsyslog.conf:/etc/rsyslog.conf
    ports:
      - '127.0.0.1:514:514'
    networks:
      - phoenix-network


  haproxy:
    container_name: 'haproxy'
    hostname: 'haproxy'
    image: haproxytech/haproxy-alpine:latest
    ports:
      - 443:443
      - 80:80
    depends_on:
      - default-backend
      - rsyslog
    volumes:
      - $PWD/haproxy-config/log:/var/log
      - $PWD/haproxy-config/certs:/usr/local/etc/haproxy/certs:ro
      - $PWD/haproxy-config/core:/usr/local/etc/haproxy/core:ro
      - $PWD/haproxy-config/maps:/usr/local/etc/haproxy/maps:ro
      - $PWD/haproxy-config/sites:/usr/local/etc/haproxy/sites:ro
    command: "haproxy -f /usr/local/etc/haproxy/core/haproxy.cfg -f /usr/local/etc/haproxy/sites/"
    networks:
      - phoenix-network

```

in `haproxy-config/core/haproxy.conf`:

```
$UDPServerAddress 127.0.0.1
$UDPServerRun 514
local0.*    /var/log/haproxy.log
& ~
```

in `haproxy-config/core/haproxy.cfg`:

```
global
  tune.ssl.default-dh-param 2048
  log rsyslog:514 local0
  maxconn 4096

defaults
  option httplog
  option forwardfor except 127.0.0.1
  option forwardfor header X-Real-IP
  option http-no-delay
  log global
  mode http

  retries 10
  option  redispatch
  timeout connect 4000
  timeout client  600000
  timeout server  600000
  timeout queue 10s

frontend default_frontend
  mode http

  bind *:80
  bind *:443 ssl crt /usr/local/etc/haproxy/certs alpn h2,http/1.1

# redirect non-www

  # http-request redirect prefix https://%[hdr(Host),regsub(^www\.,,i)] code 301 if { hdr_beg(host) -i www. }

# Make a rule that the server cannot be directly accessed by IP address

  acl   has_domain hdr(Host),map_str(/usr/local/etc/haproxy/maps/domains.map) -m found
  http-request deny if !has_domain

# ACME challenge rule?

  acl letsencrypt-acl path_beg /.well-known/acme-challenge/
  redirect code 301 scheme https if !{ ssl_fc && letsencrypt-acl }
  compression algo gzip
  compression type text/css text/html text/javascript application/javascript text/plain text/xml application/json image/svg+xml
  acl is_content_type_html res.hdr(Content-Type) -i text/html
  http-response set-header Content-Type text/html;\ charset=UTF-8 if is_content_type_html
  http-response set-header Cache-Control no-cache,\ max-age=31536000
  http-response set-header Expires %[date(3600),http_date]
  http-response set-header Strict-Transport-Security "max-age=31536000; includeSubDomains; preload"
  http-response set-header X-XSS-Protection "1; mode=block"
  http-response set-header X-Content-Type-Options "nosniff"
  http-response set-header Referrer-Policy "strict-origin-when-cross-origin"
  use_backend %[base,lower,regsub(^www\.,,i),map_beg(/usr/local/etc/haproxy/maps/sites.map,default_backend)]


listen stats
  bind *:9999
  mode http
  log global
  stats enable
  stats realm Haproxy\ Statistics
  stats uri /haproxy_stats
  stats hide-version


backend haproxy_stats_backend
  http-request auth realm haproxy-stats unless { http_auth_group(basic-auth-list) is-haproxy-stats }
  mode http
  compression algo gzip
  compression offload
  server server_nginx localhost:9999


userlist basic-auth-list
  group is-guest
  group is-haproxy-stats

  user guest password $5$N7CpS0mo$FyJtlwQOwzAi5HnCISumyBKWyPu6DhBO7eGzUUyWoJ7 groups is-guest
```

... in `haproxy-config/sites/web-push.anomalistlabs.com.cfg` ...

define an HAProxy backend and map it to the Docker host and port
NOTE:  this also turned off CORS origin rule
```
backend web_push_backend
  mode http
  compression algo gzip
  compression offload
  http-response set-header Access-Control-Allow-Origin "*"
  server server_nginx endorser-push-server:3000
```

... in `haproxy-config/maps/domains.map` ...

add a domain that will be used as a base

```
timesafari.anomalistlabs.com
```


... in `haproxy-config/maps/sites.map` ...

map the `/web-push` path to the `web_push_backend`
NOTE: `timesafari-pwa.anomalistlabs.com` PWA sits on the root

```
timesafari-pwa.anomalistlabs.com/web-push/ web_push_backend
```

### The rest ..

`docker-compose up -d` should just work :-)








## Run the server outside Docker

Run the app:

```commandline
sh <(curl https://pkgx.sh) +python.org +virtualenv.pypa.io sh

# first time
python -m venv .

source bin/activate

# first time
pip install -r requirements.txt

cp data/webpush.db.empty data/webpush.db

# For DB access, you'll have to uncomment the local path for `db_uri`.

# 3 workers would trigger 3 daily subscription runs
gunicorn -b 0.0.0.0:3000 --log-level=debug --workers=1 app:app

```

... and see the results in a browser: http://localhost:3000/web-push/vapid

See Troubleshooting below if that doesn't work out of the box.





Run a test:

```commandline
python webpush.py
```



Run haproxy (on a Mac):

* Create "haproxy-config" directory for those files above, eg. /usr/local/etc/haproxy

* Comment out the `log rsyslog` and `bind *:443` lines in /usr/local/etc/haproxy/haproxy.cfg and then run:

`haproxy -f /usr/local/etc/haproxy/haproxy.cfg`



Troubleshooting

* If you get "no such table: vapid_key" then your file pointers are probably wrong.
  Check that the "docker run" mounted volume matches the SQLALCHEMY_DATABASE_URI in the app.py file.

* If you get "unable to open database file", you can provide the app.py with
  SQLALCHEMY_DATABASE_URI with `sqlite:////...` with the full path to the data/webpush.db file.
  (Why does the relative path of `sqlite:///...` not work for a relative path?)

* Another potential problem with "unable to open database file" is the
  permissions on the directory or file with the DB, as set on the local volume
  that matches the docker `/app/instance/data` directory. Note that the user id
  in the Dockerfile is set to 1000; the `id` command will tell you the uid for
  your user, which should match. Then, user with uid 1000 should have write
  permissions to both the file and the directory, and have execute permissions
  to the whole path to that directory.