Jump to content

Admin tools development/SUL Audit/Old

From mediawiki.org

Old things from Admin tools development/SUL Audit that are no longer accurate/applicable.

HW Needs

[edit]

Doing this will require some fairly intensive DB querying. Because of the cross db-host nature of querying local.user against centralauth.*, we have developed the following plan to consolidate all user tables on the same server for the purpose of the audit. Each of these options allows us to run very intensive queries without having a detrimental effect on site performance as well as giving us the ability to execute the queries in SQL without having to connect to multiple hosts programmatically.

Option 1 - Replicate all clusters to a single slave

[edit]

Unfortunately, it would appear that it is not possible to slave s[1-7] to a single slave without running multiple instances of MySQL, which would negate the benefit. If, however, some method of accomplishing this is discovered, this plan would solve almost all temporal issues (replag being the only remaining issue).

Yes Done - Option 2 - Replicate s7/centralauth and script the rest

[edit]

Because some of the required tables in centralauth do no use an auto_increment as the primary key, we would slave at least the centralauth db from s7 to the audit db host. For each individual wiki we would batch insert programmatically, using the auto_increment ids in the user table to periodically update the tables. We would also write a script to even less frequently update all fields in the tables so that we grab updates to fields such as user_touched, user_email_authenticated, and edit_count.

This option does suffer from the fact that all databases are not constantly updated and that they are not all in sync for a given time since the scripts take time to execute. At least for now, these concerns are not huge as we are looking for ballpark statistics.