Ruby reaper incorrectly checks active jobs — removes every active lock as result

[markiz]

Describe the bug
https://github.com/mhenrixon/sidekiq-unique-jobs/blob/master/lib/sidekiq_unique_jobs/orphans/ruby_reaper.rb#L120-L132

This method checks sidekiq process list, and not active jobs.

Process hash looks like this in our setup:

{"hostname"=>"dss-queue-single2", "started_at"=>1591895424.547636, "pid"=>13767, "tag"=>"sg_dss_portal", "concurrency"=>5, "queues"=>["heavy"], "labels"=>[], "identity"=>"dss-queue-single2:13767:db7bbfe1a54e"}

In sidekiq dashboard it looks like this:

Note: one the db7bbfe1a54e is not the job id, but rather process id.

This results in reaper removing any locks for the jobs being actively executed, so you might see however many concurrent heavy jobs at the same time.

[markiz]

https://github.com/mperham/sidekiq/blob/master/lib/sidekiq/api.rb#L925 I think what should have been used is workers instead of info, and iterate over all jobs. Or maybe use the Sidekiq API itself.

[mhenrixon]

Released as v7.0.0.beta22

[markiz]

@mhenrixon thank you 🙏

[mhenrixon]

@markiz Thank you for the wonderfully detailed bug reports. Made it a breeze to fix

[godric]

Hi @mhenrixon, not sure if I should have opened a separate issue..

This bug is mentioned as "fixed" in the changelog for v6.0.23, but there isn't anything about it in the corresponding diff

Am I correctly assuming that the Changelog is wrong and this bug still present on the 6.x branch?
If so, is there any plans to backport the fix in 6.x ? 🙏

[mhenrixon]

Hi @godric,

Unfortunately, there will be no reaping of old jobs in v6. I am having some trouble with the changelog generator.