TTR documentation error

[pnabutovsky]

Looking at the Backburner README doc:

queue_respond_timeout 300 # number of seconds before job times out, 0 to avoid timeout

Looking at the Beanstalk protocol doc:

   The minimum ttr is 1. If the client sends 0, the server will silently
   increase the ttr to 1.

[nesquena]

Have you verified this? Was added here https://github.com/nesquena/backburner/commit/1b06f8e0384129c24abb5bb3f30b914805e018d3 https://github.com/nesquena/backburner/commit/4e7ef381709c5d892b14b64e1ba4df1739610041 by @zacviandier. If so, can you send a PR for the readme?

[pnabutovsky]

Unfortunately the current job execution code is:

 timeout_job_after(task.ttr > 1 ? task.ttr - 1 : task.ttr) { job_class.perform(*args) }

[zacviandier]

Effectively a change was made in https://github.com/nesquena/backburner/commit/457bfdfaac4e8b3dd6a49ea9ecd3a9cbcdaf0fb4, but anyway the goal here is to send -1 to avoid timeout.

Beanstalk protocol doc:

A positive value of timeout will limit the amount of
time the client will block on the reserve request until a job becomes
available.

I also made some test with different version of Beanstalkd and it doesn't react the same way every time. With 1.4.6-5: works really well but with 1.10 this create a infinite loop where Beanstalkd try to return 1 every time

[ryanjohns]

I came across this issue when trying to implement a job without a timeout and I don't think it's possible with beanstalkd.

I think the confusion here is that you guys are talking about different timeouts. @pnabutovsky is referring to the job TTR which is how long beanstalkd gives the worker that reserved the job to either release, delete, or bury. @zacviandier is referring to the reserve timeout which is how long a client's call to reserve will wait for a job to become available.

A reserve timeout of <= 0 will cause the client to block forever waiting on a job. This makes sense and works well.

It is impossible for a job to have a TTR < 1 as this is enforced by beanstalkd. If you enqueue a job with TTR < 1 backburner will silently set it to 1 instead. So there is no way to set things up such that a job runs until completion ignoring all timeouts. In the line that @pnabutovsky mentioned above, task.ttr will always be >= 1 and therefore timeout_job_after will always raise an exception after >= 1 seconds.

Even if we modified backburner to not raise Timeout::Error, beanstalkd would automatically release the job anyway leading to it getting worked more than once. Probably not desirable for most users.

TL;DR The only way to effectively have no TTR is to set it to some very high value.