Recomputing (rebuilding) the mentor pool
For better recommendations, we must recompute, or rebuild, the mentor pool at least once a day, ideally at a time when the database has little usage. This allows new users to become mentors and removes current mentors who are no longer good mentors. If we are running multiple instances of the sifter, use the following procedure to recompute the mentor pool:
- Disable all instances of the sifter except for one,
which we will use as the master sifter instance.
Rebuilding the mentor pool uses a large amount of database resources, so it is better to use only one sifter to rebuild the mentor pool. For
the sifter instance to be used as the master, type the following:
<mentor_set_name>.disable_build_mentor_pool = false
For each sifter instance to be disabled, type the following:<mentor_set_name>.disable_build_mentor_pool = true
- Rebuild the mentor pool on the master sifter instance
by specifying a time for the recomputation to take place. Because rebuilding the mentor pool requires heavy database access, schedule it for a time when the database
has little usage. For example, the following setting is for 2 a.m.:
<mentor_set_name>.build_mentor_pool_at = 0 2 * * *
The times are interpreted as follows:
- For each sifter instance you disabled, set a time to reload the mentor pool after it has been rebuilt.
The sifter will destroy the current mentor pool and reload the pool from the mentor pool table specified in the <mentor_set_name>.mentor_pool.table setting. Usually, you should set it for about an hour after the sifter instance has been reloaded. For example, to reload the mentor pool at 3 a.m.:
<mentor_set_name>.reload_mentor_pool_at= 0 3 * * *
Parent topic: Configure the sifter for mentor selection