How can i add a SSD as Journal-DB to an CEPH-OSD on a running clusternode?
- Thread starter hape
- Start date
I am in the same situation.
So one has to remove an OSD and create it new while adding the ssd.
If I read the docs right, a 250gb ssd should be fine for a 2tb hdd, right?
And then all writes will go through the ssd, right? But reads will be from the hdd directly.
What happens, if/when the ssd dies? Will the hdd still be accessible or is the entire OSD then dead?
So one has to remove an OSD and create it new while adding the ssd.
If I read the docs right, a 250gb ssd should be fine for a 2tb hdd, right?
And then all writes will go through the ssd, right? But reads will be from the hdd directly.
What happens, if/when the ssd dies? Will the hdd still be accessible or is the entire OSD then dead?
https://forum.proxmox.com/threads/proxmox-ceph-what-happens-when-you-loose-a-journal-disk.24148/
As per above link I assume all data will go "poof" on the OSD aswell. But as its Ceph and you have other nodes that contain replicas of the data you just add the replacement SSDs in place and wipe the OSD's and recreate the OSDs with the new SSD as journals on that particular node I assume.
I have not done this in production yet but in staging area when we tested Ceph that is how it works so I assume in theory that's how it works
People may correct me if I am incorrect
As per above link I assume all data will go "poof" on the OSD aswell. But as its Ceph and you have other nodes that contain replicas of the data you just add the replacement SSDs in place and wipe the OSD's and recreate the OSDs with the new SSD as journals on that particular node I assume.
I have not done this in production yet but in staging area when we tested Ceph that is how it works so I assume in theory that's how it works
People may correct me if I am incorrect