R820 build issues.

Question

Less a question and more a record of what has been happening during an R820 build, three days in and counting. First, the hardware: Poweredge R820 with 4 2.2GHz 8-core CPUs, 256GB RAM, PERC H710 Adding to the system: 8 WD Blue 500G SATA SSD drives Ubuntu 16.04 LTS OS MegaRAID 9380-8e (driving LTO7 library and SANS DIgital JBOD of 20 6TB WD Red) Note, I've installed a similar configuration on an R710 without issue, albeit running Windows Server 2012 R2 First I install the (SSD) drives and the MegaRAID. Configure 8-disk RAID5 in PERC config. Then take my bootable Ubuntu USB stick and attempt install. System does not see the USB stick. I also note that occasionally the USB keyboard becomes non responsive, and I have to unplug it and plug it back in, then it recovers. So, I burn Ubuntu to DVD and attempt again (mind you as you read along, each reboot is about 5 minutes). Boots to disk, install performed, looks great, reboot after installation, and receive an error: attempt to read or write outside disk 'hd0' So this puts me at a grub rescue prompt, and I proceed through all the different Ubuntu/grub forums I can find and try many things, none of which work. I saw a reference to large filesystems not being well handled by Ubuntu installs so next I... ...break the RAID, and create a 500GB RAID1 and the remaining 6 drives as RAID5. Attempt again, and install seems to work again, but I receive the same 'hd0' error. I had been doing a 'BIOS' boot in setup, because UEFI did not seem to see any PERC filesystems for boot, but I tried going to UEFI anyway, and that failed to provide any results. So I figure I can be a good boy and use the Lifecycle Controller to deploy. I go into the controller, skip RAID setup, and choose SLES 11 as the OS, then manual install, then approve that I do not have SLES (I have Ubuntu) and the system loads drivers and restarts. When it gets past the point of initializing firmware interfaces the activity LEDs on the SSD mirror start to blink continuously, and the screen remains blank. I have left it like this for over an hour, and nothing comes back. So, I try again using Lifecycle controller, and this time I go into the RAID controller and configure just a 500G RAID0 with one of the drives, and (prior to boot) yanked the other 7 out so they would not delay the boot process of confuse the system. I proceed then as above with selecting SuSE as the OS of choice, drivers load, reboots, and proceeds (minutes later, after all the startup pain) to just continuously hold the LED on the single drive and again, the blank screen. I have purchased two R820 systems, and I attempted to do the install on the other system (minus MegaRAID) to see if it was an issue with the system I had been working on, but the behavior exhibited was the exact same, so guessing that this is chronic to the R820 I thought I would report this on the forum so other people do not waste days of their lives messing with this when they could be taking their kids to the zoo or doing other things that are great in life rather than sitting in a cold computer room constantly being denied. Anyway, figure my next step will be to dig out some old 73GB SAS spinners and see if this nice new hardware just prefers legacy drives. I'll update as I find out more.

DELL-Josh Cr · Answer

Hi,

14.04 LTS is the one than Canonical has validated. https://certification.ubuntu.com/hardware/201209-11629/ You may want to try the older version and see if that works. Since the drives are not validated drives it is possible that they will not work properly. You may also want to try Opensuse or centos and see if they work at all.

Raymond-CSTARS · Answer

Josh, thank you, was thinking I would have to head down that path but thought I would try one more thing, and so far it looks good...have to step it through some paces. Here is what finally worked (assuming it holds up to tomorrow's trials):

Reinserted a second SDD drive to go back to RAID1 mirror.

Changed boot from UEFI to BIOS.

Booted into the Lifecycle controller

Rebuilt RAID1

Loaded drivers for SLES 11

On reboot the Ubuntu disk booted and went through its install. What was different this time though was that it found the vdisk (created by Lifecycle controller, I am guessing?) and I had a few different options presented that were not there before. In partitioning, I added a 1G /boot to the list, which already had a large Ext4 filesystem and a swap filesystem. The 16.04 load completed, and on reboot the grub menu appeared and I was able to boot into Ubuntu. I ran updates then let it be until tomorrow...both drives of the RAID mirror are blinking like mad, so I am hoping that overnight they finish whatever craziness they are doing and tomorrow i can proceed with a reboot and, if that works, some testing.

Another odd behavior I saw before (just to get documented here, in case someone else sees similar), was that during prior attempts (as mentioned before) I would get past the Initializing firmware interfaces step and the screen would just go black indefinitely...on a whim I unplugged the VGA from the front video port and plugged it into the back one, and was then presented with a grub shell. Seemed odd, that the video would not be the same between one port and the other. So when I did this (hopefully) successful installation, I left the VGA on the rear video port.

DELL-Josh Cr · Answer

Let us know how it goes tomorrow, glad you got it working.

Raymond-CSTARS · Answer

Looking good so far, will validate the procedure on the build-up of the second R820. I configured the fast internal workspace and the large external chassis storage, and both configured fine and the system boots well afterwards. Here are some notes/stats:

Disks internal to R820 - Qty. 8 500GB Samsung 850 EVO model MZ-75E500 (thought at first I was using WD Blue, but that was another build I was doing).

RAID card - MegaRAID 9380-8e

External JBOD - Sans Digital EliteSTOR ES424X12

JBOD disks - Qty. 20 6TB WD Red WD60EFRX

Speed test results:

Write speed to root (500GB SSD RAID1 on H710) - 996 MB/s (was same in later test set to "Write through")
Write speed to /work (3TB SSD RAID0) - 1.1 GB/s (limited by H710?)
Write speed to external (100TB WD Red RAID5 - cache write) - 1.5 GB/s
Write speed to external (100TB WD Red RAID5 - cache off) - 640 MB/s
Read from root - 2.9 GB/s
Read from /work - 4.2 GB/s
Read from external - 4.3 GB/s

PowerEdge Hardware General

Was this post helpful?