Users reported a bug where Discourse failed to download MaxMind’s GeoLite2 database despite having a valid license key. pfaffman debugged the issue and found that FileHelper.download was returning nil even though the URL worked via wget inside the container. The root cause was that the download URL required following redirects, which was disabled by default (follow_redirect: false). Arkshine suggested enabling redirects, which resolved the issue for pfaffman. A fix was merged into Discourse via PR #26333. However, Firepup650 initially still experienced the error after rebuilding, prompting a check of their Discourse version. They were on tests-passed, and although the fix should have been included, they later confirmed the issue was resolved after a subsequent rebuild. The problem stemmed from MaxMind’s updated download infrastructure requiring redirect handling, which was not initially accounted for in Discourse’s rake task. The community recommended catching such errors gracefully in the future to allow builds to complete with warnings instead of failing outright.
The build failed because of this. I removed the ENV variable and rebuild without it. Then went to debugging inside the container. where I went and added the key to discourse.conf, it’s still failing.
(Thankfully I did this before telling the client that they’d given me an invalid key!)
I added a puts to the rake task to print the URL that it’s trying to download the database like this:
url =
"https://download.maxmind.com/app/geoip_download?license_key=#{GlobalSetting.maxmind_license_key}&edition_id=#{name}&suffix=tar.gz"
gz_file =
FileHelper.download(
url,
max_file_size: 100.megabytes,
tmp_file_name: "#{name}.gz",
validate_uri: false,
follow_redirect: false,
)
if gz_file.nil?
puts "no gzfile"
end
puts "Got file #{gz_file} for #{url}"
filename = File.basename(gz_file.path)
Now I get this:
root@web-only:/var/www/discourse# rake maxminddb:get
Plugin name is 'discourse-topic-voting', but plugin directory is named 'discourse-voting'
Downloading MaxMindDb's GeoLite2-City...
no gzfile
Got file for https://download.maxmind.com/app/geoip_download?license_key=VALID_KEY_IS_HERE&edition_id=GeoLite2-City&suffix=tar.gz
rake aborted!
NoMethodError: undefined method `path' for nil:NilClass (NoMethodError)
filename = File.basename(gz_file.path)
^^^^^
/var/www/discourse/lib/discourse_ip_info.rb:52:in `mmdb_download'
/var/www/discourse/lib/tasks/maxminddb.rake:9:in `block (2 levels) in <main>'
/var/www/discourse/lib/tasks/maxminddb.rake:7:in `each'
/var/www/discourse/lib/tasks/maxminddb.rake:7:in `block in <main>'
/usr/local/bin/bundle:25:in `load'
/usr/local/bin/bundle:25:in `<main>'
Tasks: TOP => maxminddb:get
(See full trace by running task with --trace)
There was another topic where the problem was that the site was blocked by Maxmind, so I tried, and successfully downloaded that URL with wget from inside the container on the server in question.
It seems to me that FileHelper.download isn’t much of a Helper after all and is silently returning nil instead of a file. The file is only 44MB, so that’s not it. When I check the full headers with wget I don’t see a redirect.
Any ideas what the deal could be?
Is there a reason not to catch this error and let a build complete with an invalid key? Maybe just print a warning?
Could you check if you use the latest Discourse version? It works for Jay; I can’t see any reasons it would not work for you, especially when it follows MaxMind’s recommendation.
Downloading MaxMindDB...
Compressing Javascript and Generating Source Maps
I, [2024-03-26T13:27:38.305597 #1] INFO -- : Terminating async processes
I, [2024-03-26T13:27:38.307312 #1] INFO -- : Sending INT to HOME=/var/lib/postgresql USER=postgres exec chpst -u postgres:postgres:ssl-cert -U postgres:postgres:ssl-cert /usr/lib/postgresql/13/bin/postmaster -D /etc/postgresql/13/main pid: 42
I, [2024-03-26T13:27:38.307891 #1] INFO -- : Sending TERM to exec chpst -u redis -U redis /usr/bin/redis-server /etc/redis/redis.conf pid: 111
111:signal-handler (1711459658) Received SIGTERM scheduling shutdown...
2024-03-26 13:27:38.307 UTC [42] LOG: received fast shutdown request
2024-03-26 13:27:38.314 UTC [42] LOG: aborting any active transactions
2024-03-26 13:27:38.321 UTC [42] LOG: background worker "logical replication launcher" (PID 51) exited with exit code 1
2024-03-26 13:27:38.322 UTC [46] LOG: shutting down
111:M 26 Mar 2024 13:27:38.353 # User requested shutdown...
111:M 26 Mar 2024 13:27:38.353 * Saving the final RDB snapshot before exiting.
2024-03-26 13:27:38.415 UTC [42] LOG: database system is shut down
111:M 26 Mar 2024 13:27:39.896 * DB saved on disk
111:M 26 Mar 2024 13:27:39.896 # Redis is now ready to exit, bye bye...
FAILED
--------------------
Pups::ExecError: cd /var/www/discourse && su discourse -c 'bundle exec rake themes:update assets:precompile' failed with return #<Process::Status: pid 3889 exit 1>
Location of failure: /usr/local/lib/ruby/gems/3.2.0/gems/pups-1.2.1/lib/pups/exec_command.rb:132:in `spawn'
exec failed with the params {"cd"=>"$home", "hook"=>"assets_precompile", "cmd"=>["su discourse -c 'bundle exec rake themes:update assets:precompile'"]}
bootstrap failed with exit code 1
** FAILED TO BOOTSTRAP ** please scroll up and look for earlier error messages, there may be more than one.
./discourse-doctor may help diagnose the problem.
fe4eec7726fdbe08aaa2691e806bb17423a8f4eaedb8414397e3097d725e04c7
As a later side note, I didn’t link to the site this occurred on, because one of the staff members was testing a kernel bug in a separate container on the same server, and crashed the whole server by mistake, so linking to the site wouldn’t have helped at all.
What is the hash version displayed? We can quickly check, at the very least.