Skytable 0.8.1: Recovery system and fixes

We’re extremely happy to announce that Skytable 0.8.1 has been released. While we initially planned to roll out a lot of new additions in this release, we realized that it’s more important to first roll out the new recovery system along with some fixes. We strongly recommend this update for all our users as it greatly improves the operational reliability of the engine.

Recovery system

In the first Octave beta, we released an auto-recover system which however had a lot of shortcomings and bugs resulting in its quick removal before the GA release. However, this meant that the system was limited in doing what it could post a situation that would need recovery of the database files stored on disk.

With this update, in the event that an unforeseen system crash occurs (such as a due to a catastrophic system failure or power outage), the administrator can now manually run skyd repair which should automatically detect and make any appropriate repairs. The previous files will always be backed up in the backups/ directory in the event that they are required for debugging or for other purposes.

It’s important to note that the recovery system is very conservative. This means that the recovery system will restore the database to an earlier (correct) state and any data post that point is deemed unreliable and not restored.

Improved reliability

We also made a bunch of miscellaneous fixes in the CLI where --eval output wasn’t directed correctly alongside several other small fixes. Overall, with this update the engine is now very well tested and the coverage of various crash scenarios has been greatly increased. For example, multiple runtime crash scenarios and on-disk corruption scenarios are now simulated, tested and accounted for. We’ll also work on documenting the repair process in depth next week. Additionally, sysctl report status has now been fixed and augmented, enabling it to correctly report system-wide health of the server.

Review of Q1 and what’s next

Overall, this was a very productive quarter; we made the stable release of Skytable Octave just earlier this month after over two months of beta testing and with bug reports from our amazing community. And just today we added critical systems for ensuring data integrity. We look forward to a very productive Q2 where we have a lot of important system implementations such as clustering and replication lined up along with several other application level additions such as new data types and real-time queries.

Go ahead and update to Skytable 0.8.1. As always, if you find any bugs or have any suggestions please open an issue and we’ll try to do the best we can to help.

We wish everyone a productive Q2! See you at the next release :)