Re: ndb_restore

[Frazer Clement]

Hi,
Thank you for offering to file some bug reports!

I agree with your points about the ndb_schema_dist_upgrade_allowed option, the default is not correct, and there should be more safeguards before the schema distribution upgraded is allowed. Work on this is underway, but that does not help you unfortunately.

This upgrade of this functionality is currently controlled by the MySQLDs as it is only used by the MySQLD layer to distribute schema changes between themselves. However I do take your point that it is not good that one MySQLD can negatively affect the service offered by others - this breaks one of the purposes of having multiple MySQLDs in the first place - isolation and redundancy.

I would say that it is slightly unusual to have MySQLDs joining the cluster for periodic tasks like you describe, but there is nothing in the architecture which prevents it or normally makes it problematic, prior to this upgrade functionality.

On the last question about upgrading the schema distribution while there are data nodes or MYSQLD nodes on the old version - I cannot think of a situation where this would be useful, and so we should make it impossible.

Sorry once again for the problems this has caused - hopefully you can get your systems stabilized again quickly.

Thank you for taking the time to reply,
Frazer