|Where||SambaXP 2016, Berlin, Germany|
CTDB needs to be cleanly restructured into some well defined components. One benefit will be a slimmer clustered database daemon, providing an opportunity for improved SMB scalability in products such as IBM's Spectrum Scale. Other benefits include maintainability and lower barrier to entry for new developers when cluster management, public IP failover and service management are more cleanly modularised. This has been previously discussed, including at SambaXP 2015.
The process of untangling different functionality is continuing slowly. I'll discuss subsystems that have been cleaned up and what still needs to happen. Amitay's work on the parallel recovery helper has provided inspiration to improve other subsystems in a similar way. For operations that are run infrequently and are not performance critical, breaking out code into a separate executable helper provides a nice way of separating code. I'll describe some of the subsystems that have been separated.
PDF (572 KB)