Unsolved
This post is more than 5 years old
2 Posts
0
1193
PERC H330: RAID 1 detected as /dev/sde
Hi all,
Apologies if this has been covered before, I scanned the forum but couldn't see anything.
I'm working on a proof of concept for an OpenStack cloud, and have 3 x PowerEdge R530's for a storage cluster. They have a PERC H440.
These servers have 2 x 300GB 10k SAS drives for the operating system, then then 4 x 1TB disks for the rest.
When I build a RAID 1 on the SAS drives, and boot into the operating system, it detects the 1TB disks as: /dev/sd{a,b,c,d} and the RAID 1 as /dev/sde - this causes problems with the automated installer (ironically, Crowbar - a former Dell project!) we're using as it will invariably want to install to /dev/sda.
I'm working with the software maintainers and have found a fairly hacky workaround - but if anyone knows of a way to alter this so virtual disks are lettered at the start rather than end I'm all ears.
Thanks,
Craig
theflash1932
9 Legend
9 Legend
•
16.3K Posts
0
September 26th, 2016 06:00
Sounds like it is enumerating non-RAID disks first. Leave the 1TB disks out until after the install, or configure the 1TB disks in a RAID array - single-disk RAID 0's if necessary.
Not familiar with the H440 - I suspect you meant the H330 or the H730. Not sure if the H330 will support that many Virtual Disks. H730 will, and should be your choice if performance is at all important.
CScaffa
2 Posts
0
September 26th, 2016 06:00
Hi theflash1932,
Thanks for the response - you're quite right it does seem to pick the non-raid disks first.
I had the same thought as you, but unfortunately as part of the boot process with Crowbar there's a "discovery" portion - if the disks aren't in, it does not put them in the chef configuration file for the next step.
My current workaround is HBA mode with software raid on the 300Gb disks which works fine and getsthrough the installer.
Wonder what happens if I flag the CEPH nodes as "RAID capable" in the RAID controller but just don't build a RAID array. Worth a shot.
Thanks,
Craig