rebuildSearchIndex, job_runner, task_runner and cron

Software Support
1

Hello All,

This week I just ran a test update of a small journal we have here, from 3.3.0-21 to 3.5.0-4. While the update itself ran fine, there are still problems with rebuilding the search index and cron jobs.

Running php tools/rebuildSearchIndex.php returned this:

Clearing index … done
Indexing “Annual Review of Biomedical Sciences” … 168 articles scheduled for indexation

As I ran it twice, now there are 284 queued PKP\jobs\submissions\UpdateSubmissionSearchJob in my queued jobs list.

Now, in my config.inc.php I have:

job_runner = Off

and

task_runner = Off

And have

/php lib/pkp/tools/scheduler.php run

Configured via cron

The output of the cron job is:

2026-02-11 16:35:01 Running [PKP\task\ProcessQueueJobs] … 4.59ms DONE

End even if it runs once every minute, no jobs are removed from the queue.

I remember that when I ran rebuildSearchIndex.php job_runner and task_runner were set no On, so maybe this is why those jobs are queued.

php lib/pkp/tools/scheduler.php list

Does not show UpdateSubmissionSearchJob on the list.

config.inc.php says:

; Use of the built-in job runner is highly discouraged for high-volume
; sites. Instead, a worker daemon or cron job should be configured
; to process jobs off the application’s main thread.

and also says

; Use of the built-in scheduled task runner is strongly discouraged for high-volume
; sites. Use your operating system’s task scheduler instead, and configure
; it to run the task scheduler every minute.

As my target is to upgrade a much, much bigger OJS portal, I would like to avoid using this things and rely only on cron, if possible.

What is the best way to clear this jobs and reindex everything? Do I really need job_runner and task_runner?

Regards,

Oberdan

2

Maybe the problem is related to this:

3

For 3.5+, to make job run via schedule task, we have introduced a new config as

; Controls whether queued jobs should be processed by the task scheduler.
; This setting has no effect when the job_runner and the [schedule].task_runner are enabled,
; on this situation the jobs will be processed solely by the job runner.
process_jobs_at_task_scheduler = Off

see at ojs/config.TEMPLATE.inc.php at stable-3_5_0 · touhidurabir/ojs · GitHub

As it’s a upgrade, this is probably not set in the config.inc.php so it needed to be added and enabled . that should do it .

but the best approach is to use the queue worker as described at https://docs.pkp.sfu.ca/admin-guide/en/deploy-jobs .

BTW, the Queue running via CLI worker does not trigger plugin hooks · Issue #9345 · pkp/pkp-lib · GitHub is a bug but pretty sure it’s not what you are facing .

Regards
PKP Dev Team

4

Hello @touhidur ,

Running

php lib/pkp/tools/jobs.php work

Did the trick. I guess I’ll have to read all the admin documentation again for this version, several things are now different.

Just one more thing: the whole process was way faster than it was in version 3.3. Does it really extracts the text of the PDFs?

Regards,

Oberdan

5

The job UpdateSubmissionSearchJob run for each submission that is passed and delegate the actions to specific action handler which is for OJS is ArticleSearchIndex, thats all the jobs does . In OJS 3.3 we did not have queue jobs so I can not compare it but unless there is any error in error log, it should run properly .

6

Yeah, it worked and is running much faster now. Thank you very much.

Regards,

Oberdan