Bugzilla – Bug 1226367
[doc] Issue in "Creating a clustered MD RAID device"
Last modified: 2024-06-21 01:01:40 UTC
Creating a clustered MD RAID device: https://documentation.suse.com/sle-ha/15-SP5/single-html/SLE-HA-administration/#sec-ha-cluster-md-create Hi, since I got a customer case about cluster MD, please improve this part: - original: For better stability, use persistent device names such as /dev/disk/by-id/DEVICE_ID instead of /dev/sdX names. - new proposal: Use must use cluster-wide persistent names, eg. /dev/disk/by-id/xxxx; note that /dev/sdX and also /dev/dm-X are not stable names through the whole cluster; doing array management operation with non-deterministic names (eg. /dev/sdX or /dev/dm-X) can lead to catagrophic consequences. Cf. https://en.opensuse.org/SDB:SCSI_Device_Identification https://www.suse.com/support/kb/doc/?id=000018449 The main message here is - dm-X is not stable too! I hope you understand; you know better wording. Thank you very much.
And, maybe a warning box for this for better visibility???
Created attachment 875492 [details] screenshot
Thanks Jiri, I agree it makes sense to be as clear as possible there! I've added it to our queue and will make sure to get to it quikcly.
What do you think of this? --- Warning: Always use persistent device names Always use cluster-wide persistent device names, such as /dev/disk/by-id/DEVICE_ID. Unstable device names like /dev/sdX or /dev/dm-X might become mismatched on different nodes, causing major problems across the cluster. --- Here's the PR: https://github.com/SUSE/doc-sleha/pull/393
(In reply to Tahlia Richardson from comment #4) > What do you think of this? > > --- > Warning: Always use persistent device names > > Always use cluster-wide persistent device names, such as > /dev/disk/by-id/DEVICE_ID. Unstable device names like /dev/sdX or /dev/dm-X > might become mismatched on different nodes, causing major problems across > the cluster. > --- > > Here's the PR: https://github.com/SUSE/doc-sleha/pull/393 cool; thank you.
This is now merged and backported, and will be published with the next weekly docs update.