Pranith Kumar Karampuri

> **Description of problem:** > We have a glusterfs volume used to store some grafana data. There is a sqlite database stored in the volume. > > After upgrading the...

@tshacked I see that the release notes is done by you for 8.5, could you help find the rpms? @MarvinTO It looks similar to https://github.com/gluster/glusterfs/issues/2443 which is fixed in 8.6...

> @tshacked I see that the release notes is done by you for 8.5, could you help find the rpms? > > @MarvinTO It looks similar to #2443 which is...

Want to clarify a bit here. Jump consistent hash works really well for volumes that are expanding. For shrinking, one should always remove-brick from the last subvolume for better performance.

While it is definitely better than the existing layout management in DHT, we should still try to come up with a solution that avoids the full filesystem scan.

> > Want to clarify a bit here. Jump consistent hash works really well for volumes that are expanding. For shrinking, one should always remove-brick from the last subvolume for...

> It really seems a very interesting way to distribute data. > > > While it is definitely better than the existing layout management in DHT, we should still try...

> I'm not sure if this could have a big impact in performance for creates and unlinks. An alternative is to scan the brick locally (much faster than scanning through...

@xhernandez I may get some time to do some more work in this regard for other hash functions you may want me to try out. Let me know. I will...

> @pranithk First of all, these numbers look very good, and everything that reduces the number of files that are being migrated is welcomed. > In the last couple of...