In Drupal Gardens, we ended up completely losing one pair of HA
database disks, forcing us to restore that one database cluster from
backup (in all other cases, we lost at most one disk from an HA pair
so no recovery from backup was necessary because we still had an
up-to-date instance). We make database backups every hour, so for that
database cluster we lost up to one hour of data (I do not know offhand
exactly how long it was from the last backup to the crash so I don't
know the exact amount of time that was lost). That database cluster
served about 1200 Drupal Gardens sites.
So, bottom line, we lost between 0 and 60 minutes of data for about
1,200 (out of over 45,000) Drupal Gardens sites.
In Drupal Gardens, we ended
In Drupal Gardens, we ended up completely losing one pair of HA database disks, forcing us to restore that one database cluster from backup (in all other cases, we lost at most one disk from an HA pair so no recovery from backup was necessary because we still had an up-to-date instance). We make database backups every hour, so for that database cluster we lost up to one hour of data (I do not know offhand exactly how long it was from the last backup to the crash so I don't know the exact amount of time that was lost). That database cluster served about 1200 Drupal Gardens sites.
So, bottom line, we lost between 0 and 60 minutes of data for about 1,200 (out of over 45,000) Drupal Gardens sites.