This commit adds a `MiniSchedulerLongRunningJobLogger` class which will
poll every 60 seconds for mini_scheduler jobs which are stuck. When it
detects that a job is stuck, it will log a warning message with the
current backtrace of the thread that is executing the job.
Note that for scheduled jobs which are executed at a frequency of less
than 30 minutes, we will log when the job has been executing for 30
minutes.
For scheduled jobs executed at a frequency of less than 2 hours, we will
log when the job has been executing for a duration greater than its
specified frequency.
For scheduled jobs executed at a frequency greater than 2 hours, we will
log as long as the job has been executing for more than 2 hours.
This commits introduces the `sidekiq_report_long_running_jobs_minutes`
global setting which allows a site administrator to log a warning in the
Rails log when a Sidekiq job has been running for too long.
The warning is logged with the backtrace of the thread that is
processing the Sidekiq job to make it easier to figure out what a
sidekiq job is stuck on.
This commit introduces our own handling and warning for Sidekiq's new 'non-json-serializable' warning. This decouples us from Sidekiq's own deprecation cycle, and allows us to use our own deprecation system. It also means that the dump/parse happens in test mode, which will help us to catch occurrences before they reach production.
Most of our logging goes through Rails.logger, and therefore appears in Logster at `/logs` on a site. The Sidekiq logger was bypassing this and writing directly to STDERR.
Unfortunately it's not possible to do `Sidekiq.logger = Rails.logger` because `Sidekiq#logger=` applies a number of patches to the logger instance, causing our whole logging system to break.
Instead, this commit adds a dedicated Logger instance with no output, which is then patched to forward all messages directly to `Rails.logger`
Sidekiq was failing to boot in dev due to the following error. It seems
like constantizing stuff before the autoloader has kicked in caused
stuff to go weird. Root cause has not been identified but there is no
reason for us to have to warm up the cache during the initialization of
Rails.
```
2021-09-06T04:28:43.338Z pid=112028 tid=26vc WARN: NameError: uninitialized constant #<Class:0x0000564b365040d8>::RateLimiter
2021-09-06T04:28:43.338Z pid=112028 tid=26vc WARN: /discourse/app/models/post.rb:9:in `<class:Post>'
/discourse/app/models/post.rb:6:in `<top (required)>'
/ruby/2.7.4/lib/ruby/gems/2.7.0/gems/zeitwerk-2.4.2/lib/zeitwerk/kernel.rb:26:in `require'
/ruby/2.7.4/lib/ruby/gems/2.7.0/gems/zeitwerk-2.4.2/lib/zeitwerk/kernel.rb:26:in `require'
```
Previously we had many places in the app that called `hostname` to get
hostname of a server. This commit replaces the pattern in 2 ways
1. We cache the result in `Discourse.os_hostname` so it is only ever called once
2. We prefer to use Socket.gethostname which avoids making a shell command
This improves performance as we are not spawning hostname processes throughout
the app lifetime
Under extreme load on large databases certain regular jobs can take quite
a while to run. We need to ensure we never starve a sidekiq from running
mini scheduler, cause without it we are unable to queue stuff such as
heartbeat jobs.
This reduces chances of errors where consumers of strings mutate inputs
and reduces memory usage of the app.
Test suite passes now, but there may be some stuff left, so we will run
a few sites on a branch prior to merging
This commit also cleans up a bunch of pointless noise each time we boot app
- narrative was loading i18n cause redefinition of consts
- discourse.rb was loaded twice as was auth
- bin/unicorn now does all the smart things and boots unicron in dev
- bin/rails s will boot unicorn with no params
- remove bin/puma which only causes confusion
Alan Guo Xiang Tan
David Taylor
Daniel Waterworth
Loïc Guitaut
Sam Saffron
Joffrey JAFFEUX
Dan Ungureanu
Krzysztof Kotlarek
Guo Xiang Tan
Neil Lalonde
Régis Hanol
Robin Ward