Hello,
I'm in the process of doubling the number of OSD nodes in my Nautilus
cluster - from 3 to 6. Based on answers receive from earlier posts to
this list, the new nodes have more NVMe that the old nodes. More to the
point, on the original nodes the amount of NVMe allocated to each OSD
was about 120GB, so the RocksDB was limited to 30GB. However, for my
workload 300GB is probably recommended.
As I prepare to lay out the NVMe on these new nodes, I'm still trying to
understand how to size the DB and WAL for my OSDs and whether Journal is
even needed.
According to https://docs.ceph.com/en/nautilus/ceph-volume/lvm/prepare/
<https://docs.ceph.com/en/nautilus/ceph-volume/lvm/prepare/>
Bluestore supports the following configurations:
* A block device, a block.wal, and a block.db device
* A block device and a block.wal device
* A block device and a block.db device
* A single block device
First question: On my first nodes I managed to get a DB, but no WAL.
My current perception is that WAL and DB occupy separate
physical/logical partitions. By specifying a WAL size and a DB size,
ceph-volume will create the corresponding logical volumes on the NVMe.
Is this correct? It is also possible to lay these out as basic logical
partitions?
Second question: How do I decide whether I need WAL, DB, or both?
Third question: Once I answer the above WAL/DB question, what are the
guidelines for sizing them?
Thanks.
-Dave
--
Dave Hall
Binghamton University
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
