Permission denied @ dir_s_mkdir - /app/tmp/pids (Errno::EACCES)

[iRonin]

It's late and I'm getting these:

/usr/local/lib/ruby/3.3.0/fileutils.rb:402:in `mkdir': Permission denied @ dir_s_mkdir - /app/tmp/pids (Errno::EACCES)

[tiredofit/discourse:3.2.2 22:48:24 /app/tmp] $ ls -al .
total 0
drwxr-xr-x 1 root      root       10 Jan 31 22:39 .
drwxr-xr-x 1 discourse discourse 784 Jan 31 22:48 ..
drwxr-xr-x 1 root      root       16 Jan 31 22:39 cache

What's the quick fix?

Other containers work fine.

[iRonin]

I went around this and just chown and chmod /app/tmp however not sure why that's an issue.

Second issue is that the example compose.yml has uncommented REDIS_PASS for the app - I believe it would be better to leave it commented out in the example so everything just works, if one misses it.

[tiredofit]

Not seeing the uncommented REDIS_PASS as you mentioned, but can add a routine to create the tmp directory before launching to solve this for you..

Try using the 3.2.3 release..

[iRonin]

I'm now stuck at 2025-02-01.08:48:37 [STARTING] ** [scheduling] [1] Starting cron

[tiredofit]

DEBUG_MODE=TRUE will help see what the last command executed was to see where the hangup is..

[iRonin]

2025-02-01.15:34:15 [DEBUG] /etc/services.available/30-sidekiq/run ** [sidekiq] Looking for existence of /tmp/.container/20-rails
+ '[' '!' -f /tmp/.container/20-rails ']'
+ sleep 1
+ print_debug 'Looking for existence of /tmp/.container/20-rails'
+ output_off
+ '[' true = true ']'
+ set +x

[iRonin]

[tiredofit/discourse:3.2.2 15:34:54 /app] $ ls -al /tmp/.container/
+ ls -al /tmp/.container/
total 64
drwxr-xr-x 1 root root 534 Feb  1 15:33 .
drwxrwxrwt 1 root root 152 Feb  1 15:33 ..
-rw-r--r-- 1 root root  53 Feb  1 15:29 00-startup-init
-rw-r--r-- 1 root root  54 Feb  1 15:29 01-timezone-init
-rw-r--r-- 1 root root  57 Feb  1 15:29 02-permissions-init
-rw-r--r-- 1 root root  50 Feb  1 15:33 03-monitoring
-rw-r--r-- 1 root root  56 Feb  1 15:29 03-monitoring-init
-rw-r--r-- 1 root root  50 Feb  1 15:33 04-scheduling
-rw-r--r-- 1 root root  56 Feb  1 15:29 04-scheduling-init
-rw-r--r-- 1 root root  85 Feb  1 15:29 05-logging-init
-rw-r--r-- 1 root root  55 Feb  1 15:29 06-messaging-init
-rw-r--r-- 1 root root  87 Feb  1 15:29 07-firewall-init
-rw-r--r-- 1 root root  81 Feb  1 15:29 10-nginx-init
-rw-r--r-- 1 root root 139 Feb  1 15:29 11-nginx-config-reload-init
-rw-r--r-- 1 root root  49 Feb  1 15:33 20-discourse
-rw-r--r-- 1 root root  55 Feb  1 15:33 20-discourse-init
-rw-r--r-- 1 root root  55 Feb  1 15:33 99-container-init
-rw-r--r-- 1 root root  55 Feb  1 15:29 container-startup
drwxr-xr-x 1 root root  20 Feb  1 15:33 cron

[iRonin]

There is an error in /etc/services.available/30-sidekiq/run - in one line there is 20-discourse in another 20-rails

[iRonin]

2025-02-01.15:40:36 [STARTING] /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq
+ echo -e '2025-02-01.15:40:36 \e[42m[STARTING]\e[49m /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq'
++ log_prefix
++ output_off
++ '[' true = true ']'
++ set +x
++ basename ./run
2025-02-01.15:40:36 [DEBUG] /etc/services.available/04-scheduling/run ** [scheduling] [process_restart] - Sleeping for 1 seconds
+ echo -e '2025-02-01.15:40:36 [STARTING] /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq'
+ output_on
+ '[' true = true ']'
++ basename ./run
+ case "$PWD" in
+ set -x
+ cd /app/
+ silent sudo -Eu discourse exec bundle exec sidekiq --concurrency 25 --environment production --logfile /data//logs//sidekiq.log
+ '[' TRUE = true ']'
+ '[' '' = true ']'
+ '[' debug = debug ']'
2025-02-01.15:40:36 [STARTING] /etc/services.available/04-scheduling/run ** [scheduling] [1] Starting cron
Plugin name is 'discourse-mermaid', but plugin directory is named 'mermaid'

[tiredofit]

Alright, we have a couple issues going on here.

20-discourse will start unicorn. If that's not starting then we have a problem.. Please confirm after the image boots that you don't have "unicorn" running in the process list.

Sidekiq not starting (30-sidekiq) can be resolved, is likely to bring in other issues when fixed then.

I'll build the sidekiq problem as 3.2.4

[iRonin]

Also spotted fatal: detected dubious ownership in repository at '/app' which is owned by discourse:discouse

[tiredofit]

Also spotted fatal: detected dubious ownership in repository at '/app' which is owned by discourse:discouse

Alot of those are just warnings regarding some weird behaviour discourse when starting looking for .git repositories and such. While annoying, can be ignored.

[iRonin]

After editing /etc/services.available/30-sidekiq/run I'm now stuck at:

2025-02-01.15:49:23 [STARTING] /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq
+ echo -e '2025-02-01.15:49:23 \e[42m[STARTING]\e[49m /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq'
++ log_prefix
++ output_off
++ '[' true = true ']'
++ set +x
++ basename ./run
+ echo -e '2025-02-01.15:49:23 [STARTING] /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq'
+ output_on
+ '[' true = true ']'
+ case "$(basename "$0")" in
++ basename ./run
+ case "$PWD" in
+ set -x
+ cd /app/
+ silent sudo -Eu discourse exec bundle exec sidekiq --concurrency 25 --environment production --logfile /data//logs//sidekiq.log
+ '[' TRUE = true ']'
+ '[' '' = true ']'
+ '[' debug = debug ']'
+ sudo -Eu discourse exec bundle exec sidekiq --concurrency 25 --environment production --logfile /data//logs//sidekiq.log

[tiredofit]

After editing /etc/services.available/30-sidekiq/run I'm now stuck at:

2025-02-01.15:49:23 [STARTING] /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq
+ echo -e '2025-02-01.15:49:23 \e[42m[STARTING]\e[49m /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq'
++ log_prefix
++ output_off
++ '[' true = true ']'
++ set +x
++ basename ./run
+ echo -e '2025-02-01.15:49:23 [STARTING] /etc/services.available/30-sidekiq/run ** [sidekiq] [1] Starting Sidekiq'
+ output_on
+ '[' true = true ']'
+ case "$(basename "$0")" in
++ basename ./run
+ case "$PWD" in
+ set -x
+ cd /app/
+ silent sudo -Eu discourse exec bundle exec sidekiq --concurrency 25 --environment production --logfile /data//logs//sidekiq.log
+ '[' TRUE = true ']'
+ '[' '' = true ']'
+ '[' debug = debug ']'
+ sudo -Eu discourse exec bundle exec sidekiq --concurrency 25 --environment production --logfile /data//logs//sidekiq.log

That's normal. it's doing what it needs to do. As I wrote up above, I don't think Sidekiq is your issue that's blocking things.

Can I get a logfile of the container startup (I don't need DEBUG_MODE=TRUE on), and once started up to where you see yourself hanging, can you get inside the container and get the output of
ps -ef

[iRonin]

I managed it to start with the following modifications:

chown discourse:discourse /app/tmp

nano -w /etc/services.available/30-sidekiq/run

mv /assets/discourse/plugins/mermaid /assets/discourse/plugins/discourse-mermaid
mv /app/plugins/mermaid /app/plugins/discourse-mermaid

Was able to login as the admin 🙏🏻

I had issues with Bitnami discourse.
I'm running it on Synology NAS through Cloudflare tunnel.

[tiredofit]

3.2.5 renames the mermaid plugin to discourse-mermaid. Give it a try when it finishes building- it should resolve all above issues.