Starting TokuDB with missing recovery log

Hello.
I’ve lost TokuDB recovery log(HDD has died) and TokuDB won’t even start complaining about missing recovery log:


/var/log/mysql/error.log:
...
[ERROR] TokuDB: Recovery log is missing (persistent environment information is present)
[ERROR] TokuDB unknown error 2
[ERROR] Plugin 'TokuDB' init function returned error.
[ERROR] Plugin 'TokuDB' registration as a STORAGE ENGINE failed.
...

According to documentation:

By the time of the crash the database was in read-only mode for about a month, so, I guess, there was nothing important in the recovery log.

I have a backup but loading it into a new database and building indexes will take days if not weeks.

Is there any way to make TokuDB work in this situation / extract data without starting / etc?

Currently, if you lose your PerconaFT recovery log, you lose your entire FT data store.

We are looking into implementing some functionality similar to th innodb_force_recovery but due to the nature and current implementation of PerconaFT and its transactional behaviour it is quite a delicate and difficult task.

There may be some hacky way to trick the PerconaFT recovery process by hand building a new recovery log but that would require some internals knowledge and info on the current state of the FT data set, but that would very likely take longer and be more problematic than restoring from a backup.


George O. Lorch III
Software Engineer, Percona