@mattab opened this Issue on January 3rd 2008 Owner

The information contained in the tables log* should be purged automatically and regularly. Keeping all time logs in a single table significantly slows down the stats logging process (mysql having to rebuild indices, selecting from this million rows table is time consuming, maintenance is hard, etc.).

The goal of this task is to provide automatic purge of the tracking logs, every day or every month and with optional backup in a yearly table (customized with UI settings).

Processing uniques over weeks / months without using logs
In #409, we will implement a cookie store mechanism based on a mysql lookup table. This table will contain enough data (idvisitor, ip, idsite, date first visit, date last visit) to process unique visitors over a week or a month

Log purge execution
The purge task would be triggered during the 'Maintenance process' (see #1184), and once a day maximum it would execute and try to purge logs for the day (or month) before.

  • read settings (purge every day? every month?)
  • ensure all archives have been processed for the day archives contained in the log tables. Weeks/months are not necessary to be processed as they are the sum of daily archives (and uniques can be processed using the cookie store rather than keeping the full visitor logs for the period)
  • purge tracker for all websites at once
  • backs up the logs into a yearly archive table
    • INSERT INTO log_X_backups_2010 SELECT * FROM log_X
    • one backup per purged table (log_visit, log_conversion, etc.)
  • record the auto increment values (result of SELECT MAX() from log_X)
  • purge logs (TRUNCATE TABLE)
  • inserts last auto increment in each table

New Super User admin settings

  • Purge traffic logs and process reports at least every day/month
  • Backup traffic logs into a backup table containing all logs for the year (table is piwik_log_visit_backup2010)?
  • Could also add admin setting for 'Keep visitor cookies for 1 year' (from #409)

A few interesting resources

@mattab commented on June 6th 2011 Owner

Attachment: Delete old logs piwik - Not yet rescheduled task
not yet rescheduler.png

@robocoder commented on February 3rd 2010 Contributor

After pruning, we can also use MySQL's "COMPRESS" on the corresponding archive tables. A side-effect is that the archive table is read-only, but that's ok if the raw visit information no longer exists to regenerate those archives.

@mattab commented on July 27th 2010 Owner

Risk of this task is high, and time is too limited. Performance are fine long term, the downside of not purging logs is of course much higher disk space usage, but this is less an issue for 1.0 than performance.

delaying to post 1.0.

@mattab commented on December 28th 2010 Owner

This should probably be done at same time as #53

@mattab commented on March 30th 2011 Owner

Note that when Purging is enabled, we should review the "Unique Visitors" processing or disable it when logs are not available for the requested date range.


@mattab commented on April 23rd 2011 Owner

Be careful when implementing this, with regards to archives that can be deleted after the fact in another ticket: #2328 - probably, in #2328, we should delete archives for days where the logs have already been deleted

@mattab commented on April 23rd 2011 Owner

When this is implemented, we should update various FAQs that explain that it is safe to delete archive tables & have them re-processed from logs. It will not be the case, if logs have been deleted, and this should be clearly stated in all FAQs and documentations.

@peterbo commented on June 3rd 2011 Contributor

(In [4856]) PrivacyManager / Delete old statistics from database; Refs #2233, #53, #5

@mattab commented on June 4th 2011 Owner

(In [4861]) Fixes #2233, Refs #5425

  • enable new plugin on upgrade
  • Display message "your changes have been saved"
  • fix link redirect without idSite by using smarty function {url ...}
@peterbo commented on June 4th 2011 Contributor

(In [4868]) Refs #2233, #53, #5

  • tweaking / optimizing / commenting
@mattab commented on June 5th 2011 Owner

Fixed, we can always open another ticket for more advanced "pruning" and to move data to "archived" tables. Great to have this feature in 1.5!!

@mattab commented on June 6th 2011 Owner

I see a potential bug: the task is set to "Not yet rescheduled" after I clicked "Yes" and clicked Save, and after reload the task is not scheduled. Thoughts?

(see attached screenshot)

@peterbo commented on June 6th 2011 Contributor

(In [4884]) Fixes #5425 - refactored the display "last run" and "next scheduled run" method. - Since the Scheduled Tasks Timetable is not immediately updated when the plugin is enabled, we can not rely on that value. Now the displayed times should be correct and we don't have to compensate the not yet set Schedule-Timetable value with the "not yet rescheduled" phrase.

@peterbo commented on June 6th 2011 Contributor

(In [4886]) Refs #5425 - Make sure, log deletion will not be triggered before calculated "next scheduled deletion" time.

@mattab commented on February 27th 2012 Owner

see follow up ticket to purge log_action #2805

@anonymous-piwik-user commented on February 3rd 2013

In [changeset:"3e371089ad99c853d639432e8865958bfd409214/puppetpiwik"]:
Make Git the default VCS for future builds.

refs #5425

This Issue was closed on February 3rd 2013
Powered by GitHub Issue Mirror