DE EN EN (Google)

Troubleshooting

Corrupted Asset Read Access

If Apache inside a docker container reads files over a certian size (e.g. 2 MB) from a CIFS file system, then the read access does not give expected results but instead different data each read, and each time corrupted data (e.g. invalid JPG). Your easydb is probably affected by this problem in case you use CIFS for your assets. To prevent this, the MMap optimization feature in Apache needs to be turned off. (Default: on).

Inside the docker container easydb-eas the following setting is required: (in /etc/apache2/sites-enabled/easydb-asset-server.conf)

EnableMMap off

Beware: Making changes inside of the container will be overwritten by updates and other container recreations.

We advise to use the following method instead:

To make this configuration persistent, put into e.g. /srv/easydb/config/eas.yml (assuming your base directory is /srv/easydb. In any case this is a path outside of the docker container.)

apache-mmap: "off"

For the correct indentation level see the list of valid configuration options.

And restart the container with:

docker restart easydb-eas

Troubleshoot EAS startup

In exceptional cases, the EAS does not start correctly. First, try to restart the EAS using the init script:

docker restart easydb-eas

To view the log messages during startup, this can be extended to:

docker restart easydb-eas; docker logs --tail 1 -f easydb-eas

For later review of the log messages you can also look into the log file eas-worker.log. In a few cases also eas-exception.log. Those are placed into /srv/easydb/eas/log/ outside of the container, assuming your base directory is /srv/easydb. Inside of the container it is usually /var/opt/easydb/log/eas (setting EAS_LOG_DIR).

Occupied OpenOffice ports (from EAS 4.2.38)

From version 4.2.38 onwards, the EAS checks whether all network ports available for the use of OpenOffice are available (see also EAS_SOFFICE_BASEPORT). Under certain circumstances, it may happen that, when the EAS is terminated, OpenOffice parts will continue to block the ports. This circumstance is now recognized at the start of the EAS and can be recognized in the log by an error message of the following type:

eas_general (32598): 2013-11-18 11: 12: 45,777: ERROR: designated port already in use:
  TCPv4 127.0.0.1:2002 - 0.0.0.0:0 (LISTEN)

Prior to version 4.2.38, this error was not detected automatically and led to creeping problems, e.g. OpenOffice processes that run under full load.

To solve the problem, please finish the OpenOffice process. You can use this process for example. With the following call to ps:

ps -edalf | grep 'soffice.bin. * Port = 2002'

After the process is finished, the EAS should restart.

Add new color profiles

Color profiles are separated in /opt/easydb/eas/eas/data/profiles by color space in the directories rgb and cmyk. New color profiles can be copied to the directories, which are then available.

The directory /opt/easydb/eas/eas/data/profiles is inside of the container easydb-eas. To add color profiles persistently, you need to map that directory as a docker volume to a directory outside of the container.

Restart all failed jobs

  1. Connect to the PostgreSQL database of the EAS.
docker exec -ti easydb-pgsql psql -U postgres eas
  1. On the PostgreSQL database of the EAS, the following can be executed:
BEGIN;

UPDATE eas.job SET job_status = 'pending' WHERE job_id IN (
  SELECT DISTINCT ON(derived_asset_id) job_id
  FROM eas.job
  JOIN eas.derived_asset USING(derived_asset_id)
  WHERE derived_asset.derived_asset_status = 'failed'
  ORDER BY derived_asset_id, job_time_created DESC
);

UPDATE eas.derived_asset
SET derived_asset_status = 'pending'
WHERE derived_asset_status = 'failed'
  AND derived_asset_id IN (
    SELECT derived_asset_id
    FROM eas.job
    WHERE job_status IN ('failed', 'pending')
  );

COMMIT;
  1. Additionally, let the rest of the easydb know of the changes:
UPDATE ez_assets
SET
  supervisor_next_check_date = NULL,
  supervisor_last_check_date = NULL,
  supervisor_check_count = 0;

You may start this at any time, but to reduce load, you can wait until there are no more pending and processing jobs shown for EAS at the server status, in section EAS-Jobs.

  1. As all this is an unusual intervention, there is no automatic integration into the elasticsearch index. It’s recreation has to be triggered manually in this case.

How to do this is covered in FAQ.

You may start this at any time, but to reduce load, you can wait until the following statement returns 0:

select count(*) from ez_assets where supervisor_next_check_date is not null;

(… to be executed in the easydb database. To find it and connect to it, also see the same FAQ entry.)