User:Woozle/Mastodon/setup/take 3 - Revision history

Woozle: /* Part 7 */ working!

2018-01-11T00:11:47Z

Part 7: working!

← Older revision		Revision as of 00:11, 11 January 2018
Line 327:		Line 327:

	No change.		No change.
	==Part 7==		==Part 7: working!==
	The fact that I was getting 500 errors from Apache but no "500 error" web page led me to wonder: why isn't the "500.html" page being served?		The fact that I was getting 500 errors from Apache but no "500 error" web page led me to wonder: why isn't the "500.html" page being served?

Line 339:		Line 339:
	After this, I now actually get a Mastodon-generated error page: "We're sorry, but something went wrong on our end."		After this, I now actually get a Mastodon-generated error page: "We're sorry, but something went wrong on our end."

	I restart the mastodon processes with systemctl, and now I get a Mastodon login page but without CSS!		I restart the mastodon processes with systemctl, and now I get a Mastodon login page (but without CSS)!

			The problem there turned out to be that Apache was using $mastodon as its home folder, when it should have been using $mastodon/public. Upon changing that, I get a proper log-in screen.

Woozle: /* Part 6 */ part 7

2018-01-10T15:06:23Z

Part 6: part 7

@@ Line 320: / Line 320: @@
 No time to log everything, but important -- execute this as user icmstdn from /home/icmstdn/apps/mastodon:
-: export SAFETY_ASSURED=1
+: <code>export SAFETY_ASSURED=1</code>
-: RAILS_ENV=production bundle exec rails db:setup
+: <code>RAILS_ENV=production bundle exec rails db:setup</code>
 Next: restart mastodon from systemd yet again, I guess.

Woozle: /* Part 6 */

2018-01-10T01:57:41Z

Part 6

← Older revision		Revision as of 01:57, 10 January 2018
Line 322:		Line 322:
	: export SAFETY_ASSURED=1		: export SAFETY_ASSURED=1
	: RAILS_ENV=production bundle exec rails db:setup		: RAILS_ENV=production bundle exec rails db:setup

			Next: restart mastodon from systemd yet again, I guess.

Woozle: /* Part 5 */

2018-01-10T01:56:49Z

Part 5

← Older revision		Revision as of 01:56, 10 January 2018
Line 316:		Line 316:

	...and the upshot of that seemed to be that I was running out of RAM, so I temporarily upped the RAM allocation for cloud4 to 4GB and tried again: [[/paste-8]]		...and the upshot of that seemed to be that I was running out of RAM, so I temporarily upped the RAM allocation for cloud4 to 4GB and tried again: [[/paste-8]]
			==Part 6==
			...and then after looking at the PostgreSQL log, I determined that Masto was trying to use the wrong user (should be icmstdn, not mastodon) and the mastodon db had not yet been created...

			No time to log everything, but important -- execute this as user icmstdn from /home/icmstdn/apps/mastodon:
			: export SAFETY_ASSURED=1
			: RAILS_ENV=production bundle exec rails db:setup

Woozle: /* Part 5 */

2018-01-09T19:09:03Z

Part 5

← Older revision		Revision as of 19:09, 9 January 2018
Line 314:		Line 314:

	I finally hit how to get the standalone command to be in production mode: [[/paste-7]]		I finally hit how to get the standalone command to be in production mode: [[/paste-7]]

			...and the upshot of that seemed to be that I was running out of RAM, so I temporarily upped the RAM allocation for cloud4 to 4GB and tried again: [[/paste-8]]

Woozle: /* Part 4 */ part 5

2018-01-09T15:58:17Z

Part 4: part 5

← Older revision		Revision as of 15:58, 9 January 2018
Line 310:		Line 310:
	* <code>root@cloud4:/home/icmstdn/apps/mastodon# /home/icmstdn/.rbenv/shims/bundle exec puma -C config/puma.rb</code>: [[/paste-5]]		* <code>root@cloud4:/home/icmstdn/apps/mastodon# /home/icmstdn/.rbenv/shims/bundle exec puma -C config/puma.rb</code>: [[/paste-5]]
	That led me to install the Puma gem, and now we have [[/paste-6]] (tried lots of things, pasted lots of output).		That led me to install the Puma gem, and now we have [[/paste-6]] (tried lots of things, pasted lots of output).
			==Part 5==
			The problem seems to be that Puma isn't starting. The problem with figuring out why is that running it "properly" seems to hide the error messages, and running the command specified by ExecStart ''directly'' seems to be running it in a different environment -- i.e. "development" -- with its own set of issues that would need to be resolved but which shouldn't need to be resolved for production purposes.

			I finally hit how to get the standalone command to be in production mode: [[/paste-7]]

Woozle: /* Part 4 */ more

2018-01-09T13:08:11Z

Part 4: more

← Older revision		Revision as of 13:08, 9 January 2018
Line 296:		Line 296:
	...but then another comment on that site suggests that maybe installing them that way is a bad idea. So maybe I should uninstall them.		...but then another comment on that site suggests that maybe installing them that way is a bad idea. So maybe I should uninstall them.

	Searching the web for "<code>status=217/USER</code>" pulled up [https://github.com/driskell/log-courier/issues/322 this page], which suggested that the problem might be that my Mastodon username has a dash in it (which it does). This caused me to take a look at those three systemd config files to see if it was easy to change that -- and I found that they all assumed the username was "mastodon", ''and'' assume that everything is in <code>/home/mastodon</code>. Maybe this is a problem? Changing it to <code>~~vbz-net~~</code> (original contents: [[/paste-2]]).		Searching the web for "<code>status=217/USER</code>" pulled up [https://github.com/driskell/log-courier/issues/322 this page], which suggested that the problem might be that my Mastodon username has a dash in it (which it does). This caused me to take a look at those three systemd config files to see if it was easy to change that -- and I found that they all assumed the username was "mastodon", ''and'' assume that everything is in <code>/home/mastodon</code>. Maybe this is a problem? Changing it to <code>icmstdn</code> (original contents: [[/paste-2]]).

			There are multiple things which need changing in each file, actually:
			: <code>User='''icmstdn'''</code> - all three
			: <code>WorkingDirectory=/home/'''icmstdn/apps/mastodon'''</code> - all three
			: <code>ExecStart=/home/'''icmstdn'''/.rbenv/shims/bundle exec sidekiq -c 5 -q default -q mailers -q pull -q push</code> - sidekiq config
			: <code>ExecStart=/home/'''icmstdn'''/.rbenv/shims/bundle exec puma -C config/puma.rb</code> - web config

			No joy after doing that, though -- some dumps:
			* <code>journalctl --unit-mastodon-*</code>: [[/paste-3]]
			* <code>journalctl -x -u mastodon-web</code>: [[/paste-4]] - why does it try to start the service more than once?
			Running the startup command directly gave some more detailed errors that may be helpful:
			* <code>root@cloud4:/home/icmstdn/apps/mastodon# /home/icmstdn/.rbenv/shims/bundle exec puma -C config/puma.rb</code>: [[/paste-5]]
			That led me to install the Puma gem, and now we have [[/paste-6]] (tried lots of things, pasted lots of output).

Woozle: /* Part 4 */ a clew

2018-01-09T02:15:44Z

Part 4: a clew

← Older revision		Revision as of 02:15, 9 January 2018
Line 295:		Line 295:
	: <code>apt install ruby-sidekiq</code> -- resulted in additional package installations: libjs-jquery-metadata libjs-jquery-tablesorter libjs-twitter-bootstrap ruby-celluloid ruby-concurrent ruby-connection-pool ruby-hitimes ruby-redis ruby-redis-namespace ruby-timers		: <code>apt install ruby-sidekiq</code> -- resulted in additional package installations: libjs-jquery-metadata libjs-jquery-tablesorter libjs-twitter-bootstrap ruby-celluloid ruby-concurrent ruby-connection-pool ruby-hitimes ruby-redis ruby-redis-namespace ruby-timers
	...but then another comment on that site suggests that maybe installing them that way is a bad idea. So maybe I should uninstall them.		...but then another comment on that site suggests that maybe installing them that way is a bad idea. So maybe I should uninstall them.

			Searching the web for "<code>status=217/USER</code>" pulled up [https://github.com/driskell/log-courier/issues/322 this page], which suggested that the problem might be that my Mastodon username has a dash in it (which it does). This caused me to take a look at those three systemd config files to see if it was easy to change that -- and I found that they all assumed the username was "mastodon", ''and'' assume that everything is in <code>/home/mastodon</code>. Maybe this is a problem? Changing it to <code>vbz-net</code> (original contents: [[/paste-2]]).

Woozle: /* Part 4 */

2018-01-09T01:51:04Z

Part 4

← Older revision		Revision as of 01:51, 9 January 2018
Line 287:		Line 287:
	You can check whether it's running or not with any of the following:		You can check whether it's running or not with any of the following:
	* <code>systemctl status mastodon-*.service</code>: combined status log for all three		* <code>systemctl status mastodon-*.service</code>: combined status log for all three
			* <code>journalctl --unit=mastodon-*</code>: another combined journal
	* <code>journalctl -x -u mastodon-web</code>		* <code>journalctl -x -u mastodon-web</code>

Woozle: /* Part 4 */

2018-01-09T01:35:53Z

Part 4

@@ Line 289: / Line 289: @@
 * <code>journalctl -x -u mastodon-web</code>
-...and it turns out it's not running:
+...and it turns out it's not running - full paste [[/paste-1|here]].
-Jan 09 00:09:27 cloud4 systemd[1]: mastodon-sidekiq.service: Service hold-off time over, scheduling restart.
+Clues [https://discourse.joinmastodon.org/t/services-mastodon-syskiq-and-mastodon-web-wont-start/597/3 here] suggested that maybe sidekiq needed to be installed from apt -- this package apparently had not been installed:
-Jan 09 00:09:27 cloud4 systemd[1]: Stopped mastodon-sidekiq.
+: <code>apt install ruby-sidekiq</code> -- resulted in additional package installations: libjs-jquery-metadata libjs-jquery-tablesorter libjs-twitter-bootstrap ruby-celluloid ruby-concurrent ruby-connection-pool ruby-hitimes ruby-redis ruby-redis-namespace ruby-timers
-Jan 09 00:09:27 cloud4 systemd[1]: mastodon-sidekiq.service: Start request repeated too quickly.
+...but then another comment on that site suggests that maybe installing them that way is a bad idea. So maybe I should uninstall them.