When using job:complete, ERR Error running script

[saucisson]

When i try to use job:complete () (or job:cancel ()) from within a job, i get the following error:

api_1       | 2016/09/24 14:03:45 [error] 46#0: [lua] qless.lua:305: call(): ERR Error running script (call to f_ad2dfd82189c92739c79fc30a2af6fd46c8a2793): @user_script:410: user_script:410: Complete(): Job 136793b7b1ab2ee200309398b62986cf is not currently running: complete , context: ngx.timer
api_1       | 2016/09/24 14:03:45 [error] 46#0: [lua] job.lua:213: complete(): ERR Error running script (call to f_ad2dfd82189c92739c79fc30a2af6fd46c8a2793): @user_script:410: user_script:410: Complete(): Job 136793b7b1ab2ee200309398b62986cf is not currently running: complete , context: ngx.timer
api_1       | 2016/09/24 14:03:45 [error] 46#0: [lua] qless.lua:305: call(): ERR Error running script (call to f_ad2dfd82189c92739c79fc30a2af6fd46c8a2793): @user_script:410: user_script:410: Complete(): Job 136793b7b1ab2ee200309398b62986cf is not currently running: complete , context: ngx.timer
api_1       | 2016/09/24 14:03:45 [error] 46#0: [lua] job.lua:213: complete(): ERR Error running script (call to f_ad2dfd82189c92739c79fc30a2af6fd46c8a2793): @user_script:410: user_script:410: Complete(): Job 136793b7b1ab2ee200309398b62986cf is not currently running: complete , context: ngx.timer

The job does not seem to be completed, as it is then executed 4 times.
The jobs are run in an openresy/lapis environment, provided by Docker image erikcw/lapis.
Jobs are run from the following code in nginx.conf:

  init_worker_by_lua_block {
    local Qless  = require "resty.qless"
    local Worker = require "resty.qless.worker"
    local worker = Worker.new (Config.redis)
    worker:start {
      interval    = 1,
      concurrency = 5,
      reserver    = "ordered",
      queues      = { "myqueue" },
    }
  }

It seems related to #6 .

[pintsized]

Can you post a more complete example to reproduce? If you are using the builtin worker, your job should return a boolean to indicate success / failure and the worker will manage the rest.

Looks to me like the job's status has not been updated to "running", which happens when the reserver calls queue:pop().

[saucisson]

You are right, the problem was that the worker was trying to call complete () instead of returning true|false. I cannot find where the documentation specified this behavior.

[pintsized]

Yeah, sorry looks like the docs are lacking. Also, I think the Ruby worker implementation might be a bit more robust in terms of dealing with the status changing before it tries to complete.

There is definitely some ambiguity here that could use clearing up. I think I decided upon jobs returning boolean status, because the Ruby version instead uses exceptions to handle much of the logic, which wasn't very lua-ish. In that process, I failed to document the differences.

I'll try to look at this properly soon, as a clear interface for completing / failing jobs is pretty important.

[pintsized]

I've made some changes to the develop branch which deal with this. We no longer need to return "true" from job.perform to indicate success, but returning "nil, err-type, err-msg" will fail the job and log the error.

Are you able to try this branch and confirm it works for you?

https://github.com/pintsized/lua-resty-qless/tree/develop

[saucisson]

It works on my project. Thanks!