mdadm raid problems on Orange Pi+ 2e


dvachon
 Share

0

Recommended Posts

Hi! I am trying to setup a software raid with 2 usb hard disk on Orange Pi+ 2e.

The raid is working well until i simulate a failure by unplugging the 2nd usb HD cable. The system become erratic and stop to respond command and/or crash. See below;;;;

 

NOTE:

latest Armbian image available Jessie legacy server, kernel 3.x

basic image + apt-get update/upgrade then apt-get install mdadm

armbian os is still on the sd card

there is nothing on that raid

 

 

* * * THE EXACTLY SAME TEST AND STEP WERE DONE ON A BANANA PI PRO (VANILLA SERVER, KERNEL 4.X) AND EVERYTHING IS WOKING FINE AND AS EXPECTED. * * * 

 

 

-----------------------------------------------------------------------------

root@orangepiplus2e:/mnt/md1# cat /proc/mdstat
Personalities : [raid1] [raid0]
md0 : active raid0 sda2[0] sdb2[1]
      3885726720 blocks super 1.2 512k chunks
 
md1 : active raid1 sda1[0] sdb1[2]
      10477568 blocks super 1.2 [2/2] [uU]
 
unused devices: <none>
 
root@orangepiplus2e:/mnt/md1# [  143.442699] usb 4-1: USB disconnect, device number 2
[  143.458421] usb 4-1.2: USB disconnect, device number 3
 
root@orangepiplus2e:/mnt/md1# mkdir bb
 
 
[  167.793412] md/raid1:md1: sdb1: rescheduling sector 12583040
[  167.820121] md/raid1:md1: redirecting sector 12583040 to other mirror: sda1
 
[  180.030013] ------------[ cut here ]------------
[  180.038924] WARNING: at kernel/watchdog.c:255 watchdog_timer_fn+0xf8/0x2a8()
[  180.040001] ---[ end trace 985143ed8ccec8b3 ]---
[  187.850013] INFO: rcu_preempt detected stalls on CPUs/tasks:
[  187.860000]  2: (0 ticks this GP) idle=cc7/140000000000000/0
[  187.860000]  (detected by 3, t=2004 jiffies)
[  187.860000] INFO: Stall ended before state dump start
[  209.456963] systemd-journald[207]: Received SIGTERM from PID 1 (systemd).
[  209.458151] systemd[1]: Starting Journal Service...
[  247.900013] INFO: rcu_preempt detected stalls on CPUs/tasks:
[  247.910000]  2: (0 ticks this GP) idle=cc7/140000000000000/0
[  247.910000]  (detected by 3, t=8009 jiffies)
[  247.910000] INFO: Stall ended before state dump start
...
--------------------------------------------------------------------------------------------
 
Why i have thoses problems only on Orange Pi+ 2e ????
 
Link to post
Share on other sites

Armbian is a community driven open source project. Do you like to contribute your code?

I noticed on my Orange Pi PC that if I unplug an USB harddisk the system becomes erratic and stops responding.

I believe it's a power issue like on the first Raspberry Pi where you can't unplug or plug in USB items without causing a reboot.

 

Have you tried connecting the disks through a powered USB hub?

Link to post
Share on other sites

Hi! Good idea.

 

I did the test by having a usb hub between the HD and the Opi. By disconnecting the HD from the usb hub, i have the exactly same problems. So i dont think the problem is electric.

 

I did also a very basic test by failing the drive via the software mdadm and i had problems also. Lost of the console prompt. So probably the system crashed also that way.

 

=============================================================

 
root@orangepiplus2e:/mnt/md1# mdadm /dev/md1 -f /dev/sdb1
 
[  756.040025] ------------[ cut here ]------------
[  756.050008] WARNING: at kernel/watchdog.c:255 watchdog_timer_fn+0xf8/0x2a8()
[  756.050008] ---[ end trace f63674c23ec9fbc6 ]---
 
===============================================================
 
So i dont understand why mdadm is not working out-of-the-box on Orange Pi+ 2e
Link to post
Share on other sites

 i dont understand why mdadm is not working out-of-the-box on Orange Pi+ 2e

 

You're not comparing different hardware but different software: Mainline kernel on the Banana Pro and a smelly Android kernel on Orange Pi Plus 2E.

 

Mainline kernel support for H3 is not entirely ready. I tested this branch successfully on OPi Plus 2E but it's not ready for end users. So just wait a few weeks/months and use then mainline/vanilla kernel (and a btrfs mirror and not mdraid -- it's 2016).

Link to post
Share on other sites

Mainline kernel support for H3 is not entirely ready. I tested this branch successfully on OPi Plus 2E but it's not ready for end users. So just wait a few weeks/months and use then mainline/vanilla kernel (and a btrfs mirror and not mdraid -- it's 2016).

How long do you think it'll be before a 4.X kernel based Armbian with the new HDMI PHY patches for H3 is compiled & ready to rip? Should we be planning for after kernel 4.9's release?

Link to post
Share on other sites

How long do you think it'll be before a 4.X kernel based Armbian with the new HDMI PHY patches for H3 is compiled & ready to rip?

 

I did one yesterday: https://github.com/igorpecovnik/lib/issues/508

 

The problem with H3 and mainline kernel is the device tree stuff. Everything that is not already mainlined and integrated into official kernel tree requires patches that might break with each and every increase of the kernel's minor version number.

 

I had some GbE enabled H3 devices in production half a year ago (with kernel 4.4-rc1 IIRC) but maintaining this stuff to work with every kernel version in between is just too much of a waste of time. Now we're happy we can rely on megous' Github repo since he does the job to rebase THS (throttling == thermal protection) and Ethernet patches but the image created with this approach are clearly marked experimental (since stuff might break anytime).

 

So in case you use our build system it should be pretty easy to get 4.9 up and running on H3 devices but if something fails you're on your own (very low priority fixing stuff with mainline kernel on H3 now, there's too much work to be done and we clearly lack community support)

Link to post
Share on other sites

Guest
This topic is now closed to further replies.
 Share

0