PogoPlug v4: Solid Green Light, no talk w/router

[ArmMuscles]

I installed Arch Linux on a new Pogoplug v4. I set up minidlna, samba, apache, php, and a few other things.

Last night I was SSHed in, trying to install mysql. I had started up mysqld and executed mysql_secure_installation. My SSH session froze. So I power cycled the Pogoplug.

Since then, no matter what order of connecting the ethernet cable, power cable, and USB thumb drive on which Arch is installed, the same thing happens: the device powers up and blinks green on the front. Once I plug in the ethernet cable, it blinks for a few seconds, then turns solid green and remains that way. My understanding is that this measn everything is OK.

However, in my router's settings, I check the LAN and the device has not requested an IP. Furthermore, a scan of every IP on my network using my laptop reveals that the Pogoplug doesn't have an IP at all (at least, it's not on the network at all).

I've swapped out ethernet cables. It's not the cable. I've plugged the thumb drive into my laptop and the thumb drive has all the files there, not corrupt.

From my (OSX) laptop, I've issued the command strings var/log/journal/*/* | grep MESSAGE= and the last line that shows up is from fourteen hours ago. So despite the intervening attempts to plug it in and restart since then, no more MESSAGE=* have shown up in /var/log/journal.

How can I go about diagnosing this problem, or do any of you have an idea what the problem might be?

Thank you.

[Sergeanter]

I have been playing with Pogo Mobile / V4 for the last couple of months and like it in general. I got RAID5, Samba and Twonky DLNA server working on it. Everything is good as long as I do not touch it. When for whatever reason I unplug it wo going through regular shutdown, it does not boot anymore. Filesystem on boot USB stick ( generic 4GB one) gets corrupted and needs fsck -f on a different machine.
It needs journal , free blocks and nodes repair. Surface check fsck -c comes clean.
For example, I included a file share from my Asus WL500GP router in fstab. It does not get mounted at boot , so I followed an advice in other forum to make it "mountable on demand". It mounted and Pogo immediately hung up with no ssh access possible.
I unplugged it and plugged back in. It blinks for a while, initializes eSATA enclosure, recognizes RAID . I can tell from the drive access lights it gets to that point. Then just silence, no ssh, no DHCP lease. I fix the USB stick in my laptop and it comes back.
I added boot drive to fstab to make sure it gets scheduled fsck once in a while but it made no difference.
Bottom line is unplugging Mobile when it runs or hung is a sure way to prevent it from booting next time.
I see messages with similar problems here and there. To compare , Asus WL500GP is very resilient and comes back after power loss every time.
What is wrong with this picture?

[moonman]

What f/s are you using? fsck won't work at boot without initd

[Sergeanter]

The stick is formated ext3. Since the box is Mobile (with SATA added) , f/s is mounted as ext3 at boot I believe. Root f/s was created following V4 install guide to a tee . At least there is some mentioning of journal in dmesg, that I know.
Array is formatted with ext4 but it should not matter, right? It is mounted later. From the blinkies I can tell it does get mounted even when I get no ssh.
Ext3 boot stick gets corrupted _often_ at power disconnect, I would say every time.
Once it got f..ed up to the point it could not be restored with alt superblock. These symptoms are also described by somebody in another thread here.
Don't you think there are way too many complaints about boot failing with f/s corrupt than there should be for a regular ext3 system?
Is this thing picky about USB sticks? On the other hand, I had similar problems with Kingston 2GB drive.
I do have RS232 to TTL converter but really reluctant to open the box.

[Sergeanter]

I have upgraded yesterday from default kernel version 3.1.10-29 to linux-kirkwood with 3.12.8-1.
It fixed f/s corruption at power loss and jumbo frame ethernet faults.
The good news is I should worry less about freezes after hard reset ( even though I bought APC UPS and it is on its way from Amazon).
The bad news is
- jumbo frames do not improve samba transfer speed ( I have tried anything from mtu 1500 to 9000 on both ends)
- samba transfer rate with mtu 1500 dropped from 17.5MB/s to 15.5MB/s after this upgrade. Pacman upgraded samba as well so I do not know whether to attribute drop in speed to kernel or samba.

[moonman]

Jumbo frames was discussed before: Anything higher than 16xx will result in hardware checksum offload disabled, so your CPU will have to do it, which is probably not a great idea.

[Sergeanter]

I had to install and configure input-event-daemon in order to be able to shut down with a button at the back in a clean manner when Pogo Mobile stops responding.
Here is my tiny /etc/input-event-daemon.conf
$this->bbcode_second_pass_code('', '
[Global]
listen = /dev/input/event0
[Keys]
EJECTCD = sync;echo heartbeat > /sys/class/leds/status\:green\:health/trigger;shutdown -h now
')
You would need to start the service afterwards
systemctl enable input-event-daemon
systemctl start input-event-daemon
It saves me from unplugging power in most cases, but when I do unplug hot it is the same deja vu all over again:
file system gets corrupted and fsck does not seem to run at the next boot. I have to fix orphaned inodes on another machine.
Moonman da man, can I just generate uinitrd using your instructions below?
http://archlinuxarm.org/forum/viewtopic.php?f=53&t=6729
If this is all it takes ,why is it not a part of standard install?

[moonman]

Those are not really my instructons, but yes you can. The problem is that you will have to modify uBoot environment to get it to load. AFAIK PPv4 doesn't have it predefined.

[Sergeanter]

Moonman,
Would it be correct to replace
$this->bbcode_second_pass_code('', '
alarm_ide=if ext2load ide 0:1 0x800000 /boot/uImage; then setenv device /dev/sda1; run alarm_args; bootm 0x800000; else setenv isDisk yes; fi
alarm_usb=if ext2load usb 0:1 0x800000 /boot/uImage; then run alarm_which; run alarm_args; bootm 0x800000; fi
')
with
EDIT: I am updating this with exact changes I put in uboot env
$this->bbcode_second_pass_code('', '
setenv alarm_ramdisk_ide 'if ext2load ide 0:1 0x01100000 /boot/uInitrd; then bootm 0x800000 0x1100000; else bootm 0x800000; fi'
setenv alarm_ramdisk_usb 'if ext2load usb 0:1 0x01100000 /boot/uInitrd; then bootm 0x800000 0x1100000; else bootm 0x800000; fi'
setenv alarm_ide 'if ext2load ide 0:1 0x800000 /boot/uImage; then setenv device /dev/sda1; run alarm_args; run alarm_ramdisk_ide; else setenv isDisk yes; fi'
setenv alarm_usb 'if ext2load usb 0:1 0x800000 /boot/uImage; then run alarm_which; run alarm_args; run alarm_ramdisk_usb; fi'
')
EDIT:
Here is a simple script I put in /boot to generate updated uInitrd
/boot/mkuInitrd.sh
$this->bbcode_second_pass_code('', '
mkinitcpio -k `uname -r` -g /boot/`uname -r`.img
mkimage -A arm -O linux -T ramdisk -C gzip -a 0x0 -e 0x0 -n uInitrd-`uname -r` -d `uname -r`.img /boot/uInitrd
')

[moonman]

*Should* work. You have the serial right? In case it doesn't you have means to restore. Just make sure you save the original variables' values.