Metadata performance and scalability (1)

• Volume Groups with many Logical Volumes need a lot of metadata space

• Metadata is atomic at the Volume Group level – the complete metadata must be rewritten even to make the tiniest change

• Change to Logical Volume granularity?
• Optionally store it as SQL?

• By default a copy of the metadata is placed on every PV.

• pvcreate --metadatacopies 0/1/2 – How many copies on the disk?

• pvchange --metadataignore y/n – Don't use this disk for metadata