Purpose of the queue¶

Jobs launched from the UI or from the CLI are pushed into queues in order to be processed in background.

One or several daemon processes have to be launched to execute the jobs.

A daemon process can only execute one job at a time. The daemon process cannot execute any other job until the end of the current job. You can launch several daemon processes to execute multiple jobs in parallel.

Also, the daemon processes could be run on several instance of the PIM, using the same MySQL database.

These queues allow horizontal scalability of the PIM. Therefore, you can configure servers dedicated to the execution of the jobs.

Jobs are categorized in three different types and are consumed by three queues (one for each type):

• ui_job for jobs that are launched specifically by the PIM users (except imports and exports). For instance the mass edit or mass delete jobs.

• import_export_job for import and export jobs

• data_maintenance_job for all other jobs that are launched in background

Having these 3 types allows to define a priority between jobs. For instance the following command consumes the UI jobs first, then the import/export jobs, and finally the data maintenance jobs (this is the recommended usage):

1$ /path/to/php /path/to/your/pim/bin/console messenger:consume ui_job import_export_job data_maintenance_job --env=prod

You can also run the daemon and specify how many jobs you want to execute thanks to the limit option. This is useful for development purpose.

1# Run one job then exit
2$ /path/to/php /path/to/your/pim/bin/console messenger:consume ui_job import_export_job data_maintenance_job --env=prod --limit=1

Another possibility is to launch several daemons that will consume or exclude specific job types. This could be useful if for instance too many imports/exports are launched too often. You can add one or a few additional daemons that consume import/export jobs only, and speed up the queue consumption.

1# This daemon will consume import/export jobs only.
2$ /path/to/php /path/to/your/pim/bin/console messenger:consume import_export_job --env=prod

When daemons are running, you can stop them properly by using the following command. The daemons will wait for the end of the current running job (or don’t wait if no job is running) before ending.

1$ /path/to/php /path/to/your/pim/bin/console messenger:stop-workers --env=prod

If the consumers don’t stop after a while, please check that the --env option is the same as the one used to launch the consumer.

Logs¶

The daemon process writes logs to the standard output. Adding a -vv option increases the verbosity and allows to see logs about consumed jobs. It’s your responsibility to choose where to write the logs. For example, to write in the file /tmp/daemon_logs.log:

1$ /path/to/php /path/to/your/pim/bin/console messenger:consume ui_job import_export_job data_maintenance_job --env=prod -vv >/tmp/daemon_logs.log 2>&1

Do note that you should ensure the log rotation as well.

Option #1 - Supervisor¶

It’s strongly recommended to use a Process Control System to launch a daemon in production. This is not useful in development though.

In this documentation, we will describe how to configure the Process Control System supervisor, to run a daemon process. These instructions are valid for Debian 9 and Ubuntu 16.04.

1$ apt update
2$ apt install supervisor

1[program:akeneo_queue_daemon]
2command=/path/to/php /path/to/your/pim/bin/console messenger:consume ui_job import_export_job data_maintenance_job --env=prod -vv
3autostart=false
4autorestart=true
5stderr_logfile=/var/log/akeneo_daemon.err.log
6stdout_logfile=/var/log/akeneo_daemon.out.log
7user=my_user

1$ supervisorctl reread
2$ supervisorctl update

1$ supervisorctl start akeneo_queue_daemon

Option #2 - systemd¶

If you prefer, you can use systemd, which allows multiple daemons to run at the same time, log management, and auto restart in case of failure.

As of 3.1, job consumers can be assigned specific job types they will support. This can be leveraged to make sure certain types of jobs will always be processed by a given consumer without being impacted by regular activity on the PIM.

 1[Unit]
 2Description=Akeneo PIM Job Queue Service (~/.systemd/pim_job_queue/%i.conf)
 3
 4[Service]
 5Type=forking
 6User=root
 7WorkingDirectory=/path/to/home/user/.systemd
 8ExecStart=/usr/local/bin/pim_job_queue_launcher.sh %i
 9After=apache2.service
10Restart=always
11
12[Install]
13WantedBy=multi-user.target

 1QUEUE_IDENTIFIER=${1}
 2
 3JOB_TYPES=""
 4CONF_FILE=/path/to/home/user/.systemd/pim_job_queue/${QUEUE_IDENTIFIER}.conf
 5
 6if [ -f ${CONF_FILE} ]; then
 7while read job; do
 8    JOB_TYPES+="$job "
 9done <${CONF_FILE}
10fi
11if [ -z "${JOB_TYPES// }" ]; then
12    echo "${CONF_FILE} does not exist or is empty, this consumer will support all job types"
13    JOB_TYPES="ui_job import_export_job data_maintenance_job"
14fi
15
16su -c "/path/to/akeneo/bin/console messenger:consume --env=prod ${JOB_TYPES} &" akeneo
17
18exit 0

1ui_job
2import_export_job

 1# use * if you want the operation to apply on all services.
 2systemctl [start|stop|restart|status] pim_job_queue@*
 3
 4# start a pim job queue, configuration in /path/to/home/user/.systemd/pim_job_queue/1.conf
 5systemctl start pim_job_queue@1
 6
 7# start another one, configuration in /path/to/home/user/.systemd/pim_job_queue/2.conf
 8systemctl start pim_job_queue@2
 9
10# check the logs in real time for daemon #2
11journalctl --unit=pim_job_queue@2 -f

1apt install sudo
2visudo

1akeneo ALL=(root) NOPASSWD: /bin/systemctl start pim_job_queue@*
2akeneo ALL=(root) NOPASSWD: /bin/systemctl stop pim_job_queue@*
3akeneo ALL=(root) NOPASSWD: /bin/systemctl status pim_job_queue@*
4akeneo ALL=(root) NOPASSWD: /bin/systemctl restart pim_job_queue@*
5akeneo ALL=(root) NOPASSWD: /bin/systemctl reload pim_job_queue@*