Anyone already exprience with OMV and the new HP Microserver Gen10?
HP microserver Gen10
-
-
Tried to install OMV on a gen10 performance model.
Cannot view or configure RAID volumes.
Perhaps the Marvell 88SE9230 controller is not supported by default ? -
Tried to install OMV on a gen10 performance model.
Cannot view or configure RAID volumes.
Perhaps the Marvell 88SE9230 controller is not supported by default ?hmm, this is all I could come up with. https://bugzilla.kernel.org/show_bug.cgi?id=42679
There used to be a guy who worked for HP posting here, not sure where he is lately.... Maybe he will drop by if I tag him... @Spy Alelo
-
-
Just like Batman and his batsignal lol.
That Marvell chipset has been supported on Debian since 2013, but the RAID management is for Windows only.
I'll be honest though, we don't have any of those servers in the lab since they are pretty much standard and off-the-shelf components, so we have little to do with the engineering of them and I don't know much about this particular model.
-
After a second attempt, OMV is running like a charm with software RAID on my brand new Gen10.
-
May you describe your 2nd approach? What was the difference that let you use SoftRAID successfully?
-
-
The first time I created a RAID volume in the integrated manager on the ROM, it doesn't work.
The out of the box installation works perfectly from a USB flash drive, after enabling Legacy mode in the Aptio settings.
I just created a RAID 5 volume in the OMV raid management section, and created an ext4 file system in the OMV file system section.
Finally added users and shared folders, and enabled SMB/CIFS service. -
Great to read that it is working.
Marvell chipset seems to behave similar like the B120i in the gen8 (not seeing raid configs) -
I had lot of issues trying to install omv3 on my HPE proliant microserver gen10.
After dozen of tries the only method that worked to have omv3 installed
on the ssd on the internal sata was to make a debian jessie bootable usb
and once installed the base system add the omv3 repo by hand and
install omv3 on it..Now..to be honest I'm not very happy with this microserver so far, so I
hope the experience from now on will be better, or i'll think about to
get a refund for this server unfortunately.Anyway I have two major issues right now I hope that @Spy Alelo or someone else that had same issues can help me out.
- I'm getting this issue on boot: sp5100_tco: I\O address 0x0cd6 already in use, it only happens with omv3 backported kernel 4.9.0 (doesn't seem to happen with 3.9.0 kernel)
- I noticed there is a bios firmare upgrade for gen10 on the hpe site, but unfortunately download something simple like a firmware bios seems to require some sort of magic first the site asked for register after that doesn't let you still download it, seems like you have to pay for some sort of license for it?!? (if it's not is not really clear how to get those). So what is the right way to update the firmware on the microserver
- I looked at the uefi\bios settings doesn't seem to be present a function to power on on power failure, nor a wake on alarm setting..I tried anyway to plug off the cord from the server and "make" a power failure and then plug back in, it seems it auto start up again, so I guess this function is autoimplemented even if there is no setting in the bios am I right?
- Also what about ACHI support setting in the bios for the ssd instead of RAID (that I don't use anyway) in the bios? I don't see it either
-
-
I got the entry version of the Microserver Gen10 (873830-421, X3216 + 8GB) the previous week and spent the weekend testing and installing it. My goal was/is to get a cheap x86 NAS with lots of RAM and good performance, and the cheapest pre-built solution (Thecus N4810) fitting my criteria was about 430€ - the Microserver was 200€ cheaper....
I used Debian "stretch" minimal-expert graphical installer from USB as a base for OMV 4.0.
There are a few caveats with this system, but nothing dramatic:
- currently you must disable IOMMU in Bios (its even disabled by default) because the Marvell 88SE9230 has an PCIe bug which isn't (yet) worked around in the Linux kernel (Bug). With IOMMU enabled, no drives are detected at all after the kernel initializes the controller. The single internal AMD SATA port always works fine and is ideal for an boot drive.
- the Opteron X3216 integrated GPU requires an really current kernel (4.12 from stretch backports, mainline builds 4.13 from ubuntu are OK) and non-free firmware, otherwise you get a static screen after mode switching - system works fine afterwards. Without Firmware, not even VESA works: no SystemRescueCD in graphic mode etc. Console without kms (efifb??) worked fine in Kernel 4.9.
- no sensors except k10temp are working or even detected
- I tried installing the server in my attic and discovered that HP really means "operating temp: 10-35°C": The system assumes an error if the temperature falls below 10°C and shuts down via ACPI. This caused no end of confusion....
- the sp5100_tco is IMHO a false positive, but since the system
- switches video modes
- initializes the Marvell controller
afterwards, people think the system crashes. In fact Linux continues to work, but has either no disk or no video - SSH worked for me.
Also, a few general issues i found:
- BIOS initialization takes forever, even with UEFI network stack disabled
- the fan(s) are quite noisy even on the lowest settings (no problem for me)
- the PSU in standby or low power operation has a really nasty power factor and measuring standby power consumption requires a good meter - my cheap meter went crazy and shows about 10-15W more than is really consumed
- The BIOS has the Marvell "option rom" integrated and there is no setting to disable it. This means you can boot from the Marvell SATA controller, but unfortunately you can't add another Marvell based PCIe card to the system - my system stalled booting as soon as i inserted an eSATA card. I removed the option-ROM from the additional card, without it it cannot be seen in BIOS and the system continues to boot.
- No serial port, but Linux finds an integrated tty. I found some unpopulated headers on the mainboard, I'm guessing one of the these is a UART, have not checked (yet).
- The system unfortunately comes with one single-rank 8GB DDR4 ECC UDIMM installed. This is not a good idea performance wise, at least with DDR3 the AMD APUs wanted dual-rank modules and if possible dual-channel for optimum performance.
- The chassis wastes a lot of space, if HP wanted the could have integrated 2-4 2,5" drive bays. I currently have 3 drives glued in and the 4th inside the "cold plug" bay.
- Downloading BIOS requires an account on HPE.com and you must register your model- and serial number with it. This did not work with my numbers, will try later - my unit came with the current BIOS.
I did not try the RAID modes of the Marvell controller, because it would be fake-raid anyway. I currently use plain AHCI with a few old 2,5" disks in software-raid for storage and an SSD to boot and for LVM caching.
Wake-on-LAN can be configured in the BIOS and works fine on both cards - even without enabling it in Linux, which is nice. It can be configured for each NIC separately, but the BIOS menus vanish if you disable the UEFI network stack (settings still work though).
The drive bays have an ejection lever which is nice, and hot-plugging drives seems to work like a charm, even if HP says otherwise. I'm assuming this also goes for using disks above 4TB, since there is absolutely no technical reason why such a limit should exist.All in all the system works very nice and the performance beats anything you could by for 240€, especially if you consider the 4 integrated drive bays. Some quick benchmarks of Debian 9 without OMV are here, but some test would not compile and the results of others are dodgy i think. Power consumption in my current configuration seems to be about 20-25W, which is nice for a system with SSD, 3x2,5" disks running OMV, apache and a couple of PostgreSQL databases.
-
To clarify:
- I looked at the uefi\bios settings doesn't seem to be present a function to power on on power failure, nor a wake on alarm setting..I tried anyway to plug off the cord from the server and "make" a power failure and then plug back in, it seems it auto start up again, so I guess this function is autoimplemented even if there is no setting in the bios am I right?
There is an option for that, it's in some submenu, IIRC northbridge or something. You can configure the usual "last state/power off/power on" settings and they work. I found them after looking through all the submenus, but currently have no monitor connected - can't look.Edit 2017-10-22: The Setting is Chipset -> Southbridge -> Restore on Power loss
There are currently no "wake on alarm" settings to be found (but it may still work setting it from linux, I did not try).
Quote- Also what about ACHI support setting in the bios for the ssd instead of RAID (that I don't use anyway) in the bios? I don't see it either
My system came with AHCI as default, there is nothing to configure inside the BIOS/EFI. If you want to configure the RAID controller you can download the "Marvell_BIOS_Setup_Utility_for_UEFI" without registration from HPE's website. Put it on an usb disk (or inside your EFI system partition) and run it inside the EFI shell. It's dead slow but allows for configuring the Marvell controller.
-
@qwertz123 first of all thanks for sharing your experience, I was considering to throw away the unit :°D.
But I'm definitely love the same points you mentioned in your post, so I hope to have it work good indeed.
I will try to install stretch and see if with the backport kernel will solve the nasty problem of haning login prompt at startup as you said (indeed it's only "cosmetic" as the server startup anyway and I can access through ssh and webgui as you said), indeed is what I'm experiencing right now that want me to throw the unit out of the window.
Funny thing is with basic debian jessie with stock kernel (no backports) it works well, out of the box, the problem of the sp5100_tco (that I blacklisted to solve) and the hanging at boot screen happens only if I use the backport from jessie or the stock kernel on stretch, so maybe is a bug they fixed with the new kernels, so I'm going to try your way stretch+backport kernel.
From your post I didn't understood one thing though, if you go stretch+backport kernel+non-free-firmware+omv4 on top of it will get you the fix of complete boot with vga attached to the server? I mean can you finally get it until the login prompt with a monitor atteched to the server doing this?
otherwise is just the same I've got with the situation right now (I'm trying to fix the prompt stuck before asking for login on boot)Lastly about the settings I didn't know how I missed them, now I found them thanks,...about the registration to download bios firmware is beyond stupid o.O..I created the account registered my server all is good, but still can't download the firmware it asks also to connect your account with some non more specific certificate or contract (!?!?!?!?) but the page keeps refreshing itself no download...this is F*****G annoying an really stupid, so much that I would probably need to contact the support just for this stupid thing that's really....wasting time.
Anyway one last thing qwertz did you try to issue the command dmesg --level=err ?
I didn't notice any error in my omv logs, but when I issued that command I found all of this caos with it:
Code
Display Moreroot@omv:~# dmesg --level=err,warn [ 0.000000] ACPI BIOS Warning (bug): Optional FADT field Pm2ControlBlock has valid Length but zero Address: 0 x0000000000000000/0x1 (20160831/tbfadt-658) [ 0.339779] [Firmware Bug]: ACPI: No _BQC method, cannot determine initial brightness [ 0.340364] [Firmware Bug]: ACPI: No _BQC method, cannot determine initial brightness [ 0.346278] [Firmware Bug]: ACPI: No _BQC method, cannot determine initial brightness [ 0.439043] pnp 00:04: disabling [mem 0xfeb00000-0xfeb00fff] because it overlaps 0000:00:01.0 BAR 5 [mem 0xfe b00000-0xfeb3ffff] [ 2.339819] BERT: Can't request iomem region <00000000dd91c018-00000000dd91c02b>. [ 2.501691] usb: port power management may be unreliable [ 4.261421] systemd[1]: Cannot add dependency job for unit display-manager.service, ignoring: Unit display-ma nager.service failed to load: No such file or directory. [ 4.576309] [Firmware Bug]: ACPI: No _BQC method, cannot determine initial brightness [ 4.749343] amdgpu 0000:00:01.0: firmware: failed to load amdgpu/carrizo_pfp.bin (-2) [ 4.749352] amdgpu 0000:00:01.0: Direct firmware load for amdgpu/carrizo_pfp.bin failed with error -2 [ 4.749356] amdgpu 0000:00:01.0: gfx8: Failed to load firmware "amdgpu/carrizo_pfp.bin" [ 4.749417] [drm:gfx_v8_0_sw_init [amdgpu]] *ERROR* Failed to load gfx firmware! [ 4.749440] [drm:amdgpu_device_init [amdgpu]] *ERROR* sw_init of IP block <gfx_v8_0> failed -2 [ 4.749444] amdgpu 0000:00:01.0: amdgpu_init failed [ 4.749776] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.749802] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.749829] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.749853] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.749879] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.749903] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.749935] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.749963] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.749990] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750018] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750046] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750073] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750105] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.750132] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750162] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750191] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750221] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750250] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750281] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.750309] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750338] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750365] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750394] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750421] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750453] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.750480] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750508] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750536] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750564] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750592] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750623] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.750650] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750679] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750706] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750734] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750761] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.761070] amdgpu 0000:00:01.0: Fatal error during GPU init [ 4.761083] [TTM] Memory type 2 has not been initialized [ 4.762994] amdgpu: probe of 0000:00:01.0 failed with error -2 [ 4.902348] FAT-fs (sda1): Volume was not properly unmounted. Some data may be corrupt. Please run fsck.
As you can see even a warning about the ssd where the os is installed screaming about corruption...could you check it aswell and see if you get some nasty stuff too? -
-
@qwertz123
Funny thing is with basic debian jessie with stock kernel (no backports) it works well, out of the box, the problem of the sp5100_tco (that I blacklisted to solve) and the hanging at boot screen happens only if I use the backport from jessie or the stock kernel on stretch, so maybe is a bug they fixed with the new kernels, so I'm going to try your way stretch+backport kernel.AFAIK this is because the newer kernels include/enable GPU drivers for carizzo (3rd gen GCN), but jessie does not contain the firmware needed by those drivers. Without firmware loaded you get the stuck screen. Older kernels do not have any support, so you get some hardware independent frame buffer (efifb?).
QuoteFrom your post I didn't understood one thing though, if you go stretch+backport kernel+non-free-firmware+omv4 on top of it will get you the fix of complete boot with vga attached to the server?
Yes, I only tested the VGA port though since I did not have anything to connect to DP.
QuoteLastly about the settings I didn't know how I missed them, now I found them thanks,...about the registration to download bios firmware is beyond stupid o.O..I created the account registered my server all is good, but still can't download the firmware it asks also to connect your account with some non more specific certificate or contract (!?!?!?!?) but the page keeps refreshing itself no download...this is F*****G annoying an really stupid, so much that I would probably need to contact the support just for this stupid thing that's really....wasting time.
HPE only offers firmware within the warranty period, so they need to check if you are still good. I think you need to link your device with your account ("link warranties", "Garantien verknüpfen" within the menu) and enter product and serial number. But this does not accept my numbers currently, I'll need to contact them directly to see why.
QuoteAnyway one last thing qwertz did you try to issue the command dmesg --level=err ?
I didn't notice any error in my omv logs, but when I issued that command I found all of this caos with it:
Coderoot@omv:~# dmesg --level=err,warn [ 0.000000] ACPI BIOS Warning (bug): Optional FADT field Pm2ControlBlock has valid Length but zero Address: 0 x0000000000000000/0x1 (20160831/tbfadt-658) [ 0.339779] [Firmware Bug]: ACPI: No _BQC method, cannot determine initial brightness [ 0.340364] [Firmware Bug]: ACPI: No _BQC method, cannot determine initial brightness [ 0.346278] [Firmware Bug]: ACPI: No _BQC method, cannot determine initial brightness [ 0.439043] pnp 00:04: disabling [mem 0xfeb00000-0xfeb00fff] because it overlaps 0000:00:01.0 BAR 5 [mem 0xfe b00000-0xfeb3ffff] [ 2.339819] BERT: Can't request iomem region <00000000dd91c018-00000000dd91c02b>.
These are some inconsistencies inside BIOS / ACPI tables, you can ignore that at least as far as I know. The brightness thing seems to indicate the system wants to be a laptop with integrated screen
They may vanish without a display connected or with a newer kernel, since I'm not getting them ATM.Code[ 4.749343] amdgpu 0000:00:01.0: firmware: failed to load amdgpu/carrizo_pfp.bin (-2) [ 4.749352] amdgpu 0000:00:01.0: Direct firmware load for amdgpu/carrizo_pfp.bin failed with error -2 [ 4.749356] amdgpu 0000:00:01.0: gfx8: Failed to load firmware "amdgpu/carrizo_pfp.bin" [ 4.749440] [drm:amdgpu_device_init [amdgpu]] *ERROR* sw_init of IP block <gfx_v8_0> failed -2 [ 4.749444] amdgpu 0000:00:01.0: amdgpu_init failed [ 4.749417] [drm:gfx_v8_0_sw_init [amdgpu]] *ERROR* Failed to load gfx firmware!
You're missing the firmware, so no display / kernel mode setting
Code
Display More[ 4.749440] [drm:amdgpu_device_init [amdgpu]] *ERROR* sw_init of IP block <gfx_v8_0> failed -2 [ 4.749444] amdgpu 0000:00:01.0: amdgpu_init failed [ 4.749776] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.749802] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.749829] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.749853] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.749879] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.749903] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.749935] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.749963] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.749990] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750018] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750046] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750073] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750105] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.750132] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750162] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750191] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750221] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750250] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750281] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.750309] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750338] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750365] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750394] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750421] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750453] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.750480] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750508] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750536] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750564] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750592] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750623] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 3 [ 4.750650] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750679] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 4 [ 4.750706] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.750734] [drm:dce_v11_0_set_pageflip_irq_state [amdgpu]] *ERROR* invalid pageflip crtc 5 [ 4.750761] [drm:amdgpu_irq_disable_all [amdgpu]] *ERROR* error disabling interrupt (-22) [ 4.761070] amdgpu 0000:00:01.0: Fatal error during GPU init [ 4.761083] [TTM] Memory type 2 has not been initialized [ 4.762994] amdgpu: probe of 0000:00:01.0 failed with error -2
Noise as a consequence of no gpu firmware.
Code[ 4.902348] FAT-fs (sda1): Volume was not properly unmounted. Some data may be corrupt. Please run fsck.
You did not shut down properly. If a (fat) fs is mounted, a bit "I am mounted" is set within the file system. If the system crashes the bit does not get cleared (as opposed to a clean unmount), triggering this message (thats why it says 'may be corrupt', not 'is corrupt'). Unmount it and run fsck -f /dev/sda1.
But i suspect /dev/sda1 is mounted as /boot/efi (efi system partition - bootloader, i.e. grub) and since you're getting this message from Linux after grub was found and executed - all is fine.
My output with Debian "stretch" and Kernel 4.13.5 from Ubuntu (mainline build) looks like this [1]:
Code$ dmesg --level=err [ 13.487190] Error: Driver 'pcspkr' is already registered, aborting... [ 15.134805] kfd kfd: error getting iommu info. is the iommu enabled? [ 15.134843] kfd kfd: Error initializing iommuv2 for device (1002:9874) [ 15.136814] kfd kfd: device (1002:9874) NOT added due to errors
And these messages are to be excpected, since IOMMU is disabled in BIOS.
Note: I did try workaround documented in various forums (kernel parameter iommu=pt and enabling the IOMMU in BIOS) to avoid this, but did not fully test it. I had other troubles with my PCIe SATA at the time, but i suspect with only the internal controller it may work.
Footnotes:
[1]
It's fine to use a kernel from another OS or newer release on Debian, since a) Ubuntu is not that far removed and b) the kernel-userspace-api is stable and basically never changes. -
That's really helpful info, thanks @qwertz123 , I will try today to achieve the same result.
Yesterday I tried to install stretch just with utilities and ssh server (to access through ssh after installation and download the backport kernel) but didn't success, I unfortunately forgot that debian disable ssh access for root as default and I didn't created another user at installation time..so I will try again this morning.About the kernel, to be honest, I'm not really comfortable to install ubuntu's kernel in it, especially because I want to install omv on it and I want as less problem I could have for future upgrades (that's useful anyway for reference so thanks to have mentioned it)..but you said it will work with debian stretch backport kernel too 4.12 am I right or did I misunderstood you?!
If that the case I'm all set, I don't really need IOMMU anyway because I'm no planning any vm on it, so I don't really care if it doesn't work for my needs.
Thank you a million to have taken time and explained me my dmesg message was really helpful pal!! really really appreciate that!
At last about the firmware, I already tied the product to my id and registered the serial (that went through), but still can't download it -.-, I will contact them and see what's going on.
It will be my first time to contact support just for download from a link (I guess there is one time for everything).What about performance how did you find yourself with it? and what are your use for this microserver (I have the same model as yours by the way)
-
Yesterday I tried to install stretch just with utilities and ssh server (to access through ssh after installation and download the backport kernel)
I used the minimal/network iso and the graphical expert install, since I wanted only the minimal system. Afterwards just added the OMV repository to sources.list and installed the packages according to the wiki.
QuoteAbout the kernel, to be honest, I'm not really comfortable to install Ubuntu's kernel in it, especially because I want to install omv on it and I want as less problem I could have for future upgrades (that's useful anyway for reference so thanks to have mentioned it)..but you said it will work with debian stretch backport kernel too 4.12 am I right or did I misunderstood you?!
I used the debian backports 4.12 at first and it seemed to work. Just make sure you install the non-free firmware, package firmware-amd-graphics or even better firmware-linux. This is needed regardless of the kernel version.
QuoteIf that the case I'm all set, I don't really need IOMMU anyway because I'm no planning any vm on it, so I don't really care if it doesn't work for my needs.
I want to use the IOMMU at least in passthrough configuration, because as I understand it it is needed for full graphics / hsa support. Even if I do not have a monitor connected graphics support (UVD, VCE) / opencl acceleration would be nice for things like video transcoding. But I can wait a few kernel versions for it to work, the CPU should be fast enough for anything except .h265 or 4k/8k content.
QuoteWhat about performance how did you find yourself with it? and what are your use for this microserver (I have the same model as yours by the way)
I really wanted to buy the 4 core Microserver, but it's just way to expensive.
The entry level is still really nice I think - lots and lots of RAM from the get go and the performance of a cheap laptop. I know the performance of various Debian machines with ARM processors (Raspi, Banana Pi Router), some D2550 Atoms and lots of old AM2+ CPUs - it beats the hell out of all of them, even if it is slow compared to modern desktop hardware. Benchmarks say it even beats most 4-Core x86 Atom based NAS while consuming the same amount of power....
Also, it has full AES, virtualization and should have GPU/video support, which is a major bonus.
So after the first two weeks, I'm quite happy with the Microserver.I will upgrade it in the near future though:
- 2nd UDIMM with dual rank, 8GB (the upgrade to 16GB is cheaper than exchanging the existing DIMM against 2x4GB) for performance reasons (dual-channel RAM with 3 ranks)
- 1x PCIe based 2port SATA-Card (based on ASM1061) since my current Marvell 88SE9125 card really dislikes the integrated 88SE9230
- 1x PCIe based Multi-IO-Card (2xserial, 1x LPT) for for UART and hopefully GPIO/I²C
This is is to replace some other systems (Raspi, BPi) currently running alongside the microserver to reduce power consumption of my home. Even after all those upgrades *and* converting the unit to DC power it'll be cheaper than a pre-built 4-bay NAS with 4GB RAM...
-
-
- New update now with stretch+backports kernel+firmware-nonfree it's a bit better.
- If you want to know DP works like vga (I had a dp cable around so I tested it)
- I also blacklisted sp5100_tco to get rid of that nasty already in use error (I think doesn't create any problem to the system, despite I would have preferred some more clean solution to it)
- Also the ACPI bios bug disappeared after blacklisting sp5100_tco (I don't know why or if it's related seems odd)
Now my dmesg error level is considerably lower:
Code[ 2.253391] BERT: Can't request iomem region <00000000dd91c018-00000000dd91c02b>. [ 5.215462] kfd kfd: error getting iommu info. is the iommu enabled? [ 5.215527] kfd kfd: Error initializing iommuv2 for device (1002:9874) [ 5.215758] kfd kfd: device (1002:9874) NOT added due to errors
I found this article about kfd error https://community.amd.com/thread/170292 but the fix didn't worked out for me unfortunately.At the beginning I had also a warning about usb power management, but after a couple of reboot just disappeared
it remains just the annoying kfd iommu info, why does it try to load it if it's disabled on the bios is beyond me, I will try to look at the fix you mentioned of iommu
but at least I think it's just cosmetic, so we are good!Indeed you were really helpful @qwertz123 thanks a lot!
It is a good looking machine for the price point and consumption, I've another around, I'm considering to use it in a small office with 5\6 users to use samba\share where they work coupled with 2 data disks+ 2
disk with rsnapshot to have backup of the data disks (I'm not really a raid guy, I prefere snasphots so you can go back in time) plus teamviewer host to let them turn on their pc and work from home if needed.I hope the amd apu can handle it, it shouldn't be too much difficult for it, indeed 5 users are not too much, so I hope will handle everything.
Interesting choice with your upgrades plan, especially about the ram upgrade are you planning use zfs or just to have as much power as you could?
I love the components for the price, especially the ecc support, leaving open road to future zfs upgrade aswell.Quote from qwertz123I used the minimal/network iso and the graphical expert install, since I wanted only the minimal system. Afterwards just added the OMV repository to sources.list and installed the packages according to the wiki.
I did the same but I could only find one for kralizec in the official wiki
Anyway I think not much is changed so I went ahead and went with:Codeecho "deb http://packages.openmediavault.org/public arrakis main" > /etc/apt/sources.list.d/openmediavault.list apt-get install --allow-unauthenticated openmediavault-keyring apt update apt install postfix apt install openmediavault omv-initsystem
It looks that everything is ok, but I don't know if I need other packages for arrakis, I hope not. -
it remains just the annoying kfd iommu info, why does it try to load it if it's disabled on the bios is beyond me, I will try to look at the fix you mentioned of iommu
but at least I think it's just cosmetic, so we are good!
What it tries to load is amdkfd, which is a driver for AMDs HSA architecture which is present. AFAIK while initializing amdkfd the kernel notices that IOMMU is a) required and b) not working for the (current) iGPU. Thus, it aborts integrating the CPU and GPU and tries the next device candidate - which isn't present (no 2nd GPU), so it stays disabled (but loaded).Quote
It is a good looking machine for the price point and consumption, I've another around, I'm considering to use it in a small office with 5\6 users to use samba\share where they work coupled with 2 data disks+ 2disk with rsnapshot to have backup of the data disks (I'm not really a raid guy, I prefere snasphots so you can go back in time) plus teamviewer host to let them turn on their pc and work from home if needed.I hope the amd apu can handle it, it shouldn't be too much difficult for it, indeed 5 users are not too much, so I hope will handle everything.
They sell NAS with a quarter of the RAM and half the CPU power for double the user count, so I'm not worried
Just remember to a) backup on external drives too in case something goes horribly wrong and b) use the CIFS/SMB trashcan option, otherwise you'll go crazy restoring the users "oopsies" from backups/snapshots. In my experience it pays if the user is able to help him or herself, even if there's an IT-ish guy around.Quote
Interesting choice with your upgrades plan, especially about the ram upgrade are you planning use zfs or just to have as much power as you could?
Not planning on ZFS, for my purposes btrfs will be enough (currently on ext4 though). I'm the traditional raid+lvm kind of guy which negates most "advantages" of btrfs/zfs. In fact, OMV is my first ever GUI
The additional RAM is mostly for a) cache (application caches of databases, file system cache, stripe cache / readahead) b) zram (swap) and c) performance. I do not need more ram, but I'd like the additional performance of dual channel. "Unfortunately", 2 sticks 4GB ECC are more expensive than a single one, so I'm sticking another in and get 16GB. That the disks will stay powered off most of the time because of RAM and SSD cache is just a bonus. -
Quote from qwertz123
They sell NAS with a quarter of the RAM and half the CPU power for double the user count, so I'm not worried
Just remember to a) backup on external drives too in case something goes horribly wrong and b) use the CIFS/SMB trashcan option, otherwise you'll go crazy restoring the users "oopsies" from backups/snapshots. In my experience it pays if the user is able to help him or herself, even if there's an IT-ish guy around.Thanks about the trashcan double check!! :°D I had already bad experience with people not knowing how to use a server unfortunately :°D so it's the first thing I set :°D, so I guess we share the same pain
and already set backup plugin with 2 external usb 3.0 hard drive formatted in ntfs that point to the rsnapshot hard disk so they will aslo get a portable backup with incrementals directly on ntfs partition that can plug and play on their windows machine and recover also something that they f*cked up in the trashcan aswell :D.So I guess if now they will still fuck up they deserve a medal of honor (I hope not)
Indeed really thanks @qwertz123 for all your inputs were really useful!
Now what remain to do is fix those last omv4 errors weakref.py and invalid rpc response warning and I'm good to deploy this little buddy
-
-
and already set backup plugin with 2 external usb 3.0 hard drive formatted in ntfs that point to the rsnapshot hard disk so they will aslo get a portable backup with incrementals directly on ntfs partition that can plug and play on their windows machine and recover also something that they f*cked up in the trashcan aswell :D.
If it wasn't for the "plug and play Windows" requirement I'd advise against NTFS drives for backup. NTFS write support via fuse is heavily CPU dependent. So I suspect the backups will be quite slow, USB3 or not.
I'd recommend FAT32, but that limits the file (and disk) size too much.
If you do not expect files above 4GiB, you may format partitions above 32GB with FAT32 under Linux, windows reads those fine too - it just does not create them. YMMV. -
If it wasn't for the "plug and play Windows" requirement I'd advise against NTFS drives for backup. NTFS write support via fuse is heavily CPU dependent. So I suspect the backups will be quite slow, USB3 or not.I'd recommend FAT32, but that limits the file (and disk) size too much.
If you do not expect files above 4GiB, you may format partitions above 32GB with FAT32 under Linux, windows reads those fine too - it just does not create them. YMMV.Yep unfortunately was a request to have it windows ready, but I guess I can't use fat32, they work on big files 3d and such, (I have to ask though) so I guess the NTFS is the only viable solution, exfat is excluded because no "journaling", also I guess is the same problem with ntfs on speed, in the end if the only problem will be slowness I guess they have to wait to have everything, but I will double check the 4 gig limit with them..
I just thought go directly with ntfs because I didn't want to have any trouble with them saying, oh I have this one file but now the backup doesn't work :°D, especially because they are not really tech savy and I will not have to much time to fix every single mistake they make.
(I'm not an IT guy by profession I do this for free to help them out..I just couldn't stand what another "professional IT guy\company" did manage to give them for almost 3 time the price of this unit..was a piece of junk falling apart...literally..so I just offered to build a more clear and solid solution for them with this unit and the great omv on top of it for free, so..I guess if the drawback is a bit of slowness in the portable backup they can handle it...especially in confront of the piece of junk it will substitute)
Participate now!
Don’t have an account yet? Register yourself now and be a part of our community!