Add schema to allow further controls of tictacaae #46
Conversation
This allows: - setting of log_level for AAE Backend; - setting of log_level for TictacAAE overall (hidden); - setting of compression method (support switch to better performing zstd). Previously all AAE backends had the database ID 2 ^16 - 1. They now have an ID 2 ^ 16 + Partition Number e.g. with a ring size of 1024 the partition numbers will be 0 to 1023). This allow leveled logs in parallel mode AAE to be distinguished from leveled vnode logs, and also distinguished between vnodes.
|
As part of the change the rebuild wait/delay defaults have also been increased. There was never any science behind this initial setting. The purpose is risk mitigation, for an incredibly unlikely series of events, that might occur strung out over a long period of time (i.e. data is silently corrupted on all 3 nodes without detection, despite CRC checks, and without the data being read at any time between the first and the third corruption). On production systems, even with ephemeral disks, the frequency of data corruption is incredibly rare. So rebuilding all trees every month to protect against this is just unnecessary work. Rebuilding every 3 months seems more proportionate to the risks. The operator still has the option of choosing more aggressive defaults should they see increased corruption within their environment. |
This allows:
Previously all AAE backends had the database ID 2 ^16 - 1. They now have an ID 2 ^ 16 + Partition Number e.g. with a ring size of 1024 the partition numbers will be 0 to 1023). This allow leveled logs in parallel mode AAE to be distinguished from leveled vnode logs, and also distinguished between vnodes.