More on Global Locks

[alexcrow]

Now that MooseFS seems to have global locks (https://moosefs.com/products/overview.html) and given their code is GPL, should we expect to see this soon in LizardFS? Global locks would be great for running CTDB+Samba to avoid having to use GlusterFS or NFS for the CTDB lock area.

Also can the masters still be managed by Corosync and Pacemaker until the new HA stuff is finalised? Is there any easy recipe for this around? I've looked across the repo and all the info on master HA is very confusing and often conflicting.

One more thing - a wishlist item - Metadata sharding ;-) I know that metadata stuff is really hard to do right (see CephFS!) so it may just be a pipedream. Does open the opportunity to have billions or trillions of files/chunks without paying for an 8-socket Xeon box,,,

Thanks

Alex

[psarna]

1. Locks will probably be implemented soon
2. You could still use corosync/pacemaker for ha, especially if you know how to configure it properly. If you want an easy recipe, you can try using UCARP or keepalived for basic ha
3. We are already considering various ways of reducing metadata overhead on memory,
   you can expect more information in the near future

[borkd]

If various ideas for metadata footprint and scalability are still being considered BeeGFS (formerly FhGFS) does a pretty good job in this respect.
"FhGFS distributes metadata on a per-directory basis. This means each directory in your filesystem can be managed by a different metadata server. When a new directory is created, the system automatically picks one of the available metadata servers to be responsible for the files inside this directory. New subdirectories of this directory can be assigned to other metadata servers, so that load is balanced across all available metadata servers."
(from http://www.beegfs.com/wiki/FAQ#distributed_metadata)

[onlyjob]

@borkd, BeeGFS is proprietary so they can claim pretty much anything they want and you won't be able to verify their claims. Let's waste no time for proprietary systems, shall we?