Deployment - abrt/retrace-server GitHub Wiki

If you want to deploy retrace server for testing purposes, please skip to Deploying Testing Retrace Server

Deploying Retrace Server

Build from source and install

retrace server uses Tito to manage the build process. Tito and Meson are the main prerequisites for starting the build. The following command buids an RPM package from the most recent revision of the source code and installs it immediately:
```
 $ tito build --test --rpm --install
```
Deploy

After installing the retrace server, before having a functional retrace server you need do a few things:
1. Add repositories If you need to add another distribution, another repository or new url for mirror do it in /usr/share/retrace-server/plugins/distribution.py. To learn more about writing plugins please read Plugin section in README
2. Run reposync To download packages from repository run (must run as retrace user):
```
 $ sudo -u retrace retrace-server-reposync distribution version architecture
```
  Don't forget to substitute the last three arguments. You want to run this command on all combinations of distribution version and architecture always when new packages or new versions are releases. Therefore it is recommended to set this into cron.
3. Open port 443
  
  To enable communication via https you have to open port 443. You can do so by running these commands:
  
  Open port 443 in the firewall
```
 # firewall-cmd --permanent --zone=public --add-service=https
```
  Reload the firewall
```
 # firewall-cmd --reload
```
  Make sure that the port is listed
```
 # firewall-cmd --zone=public --list-ports
```
4. Restart httpd
```
  # systemctl restart httpd
```
5. Disable SELinux
```
  # setenforce 0
```
Test your server

There are two ways how to test if your server is running:
- Abrt Plugin
  
  There is plugin in ABRT that should be available in your system if you have ABRT installed. See abrt-retrace-client -h to learn more.
- Web UI
  
  You should be able to see front page by visiting https://my_server
  
  There also exist a Web UI for assigning retrace task. It can be found under \manager, but firstly must be enabled in configuration file.
```
  open /etc/retrace-server/retrace-server.conf

  set `AllowTaskManager = 1`
```
  Mind that this a security risk and should not be enabled in production servers.

Deploying testing Retrace Server

When deploying retrace server only for testing purposes, you don't want to have all packages from all repositories. You only need some packages, so you can retrace one or two coredumps. In this section a simple tutorial is written how to deploy such a server. Each point corresponds with point from section Deploying Retrace Server.

Build from source and install

There is no change when deploying real and testing retrace server.
Deploy
1. Add repositories
  
  This is the main difference. Since you don't want to have all packages from fedora, you should NOT run commands like $ retrace-server-reposync fedora version architecture You only need a few packages (and their dependencies). Therefore you should create a local repository, download only necessary packages there and tell retrace-server that this is your repository. Here is how you can do it.
  1. Create a new folder and call it for example local_repo
```
 $ mkdir /var/tmp/local_repo
```
  2. Now you have to download packages into that folder. If you have yours coredump, find out which package it comes from - it will be later marked as my-packages.
  Run this:
```
 # dnf --releasever=31 --enablerepo=\*debuginfo\* -y --installroot=/var/tmp/local_repo/ \
 download --resolve --destdir /var/tmp/local_repo/ abrt-addon-ccpp shadow-utils \
 gdb rpm *my-packages*
```
  Then if you look into /var/tmp/local_repo you should see a few packages.
  
  It is your own local repository. Now you need to create a plugin for it. Easiest way is to copy existing plugin and create changes in it.
```
 $ cd /usr/share/retrace-server/plugins/
 # cp fedora.py local.py
```
  Now open /usr/share/retrace-server/plugins/local.py in your favourite text editor and do the following changes: All occurances of word Fedora replace by Local, all fedora by local and you only want one repository with one mirror, so your repos part should look like this:
```
    repos = [
        [
            "/var/tmp/local_repo"
        ]
    ]
```
2. Run reposync
```
 $ sudo -u retrace retrace-server-reposync local 31 x86_64
```
3. Open port 443
  
  There is no change when deploying real and testing retrace server.
4. Restart httpd
  
  There is no change when deploying real and testing retrace server.
5. Disable SELinux
  
  There is no change when deploying real and testing retrace server.
Test your server

You should be able to test your server the same way as when server deployed normally. Only difference would be, that you cannot update just any coredump. The package from which the crashing app comes must be in your repository. You can download it later and run reposync again and you should be good to go. And do not forget, your release is now called Local release 31 (Thirty one).

Frequent problems when testing a retrace server deployment

Submitting coredump via manager from local folder and retrace server cannot find it

Create tarball of the coredump tar -cf coredump.tar coredump and put it into /var or /var/spool. Then put path as file:///var/(spool/)coredump.tar

Task fails on running mock (last command starts with /usr/bin/mock init --resultdir...)

There is a lot of possibilites, but most often it is one of two:

* You have not enough space in `/usr/lib/mock`

* A gpg check failed. Set `RequireGPGCheck = 0` in `/etc/retrace-server/retrace-server.conf`
  and restart httpd.