No space left on device

adukes40

@george1421 when I type fog…it says…An error has been detected…fatal error unknown request type null

george1421

@adukes40 did you schedule a debug deploy? I’ve only seen this error before when you use the USB boot stick. The fog server should set the type kernel parameter.

adukes40

@george1421 I got in earlier. I did a normal debug, and not a debug deploy… I am trying to putty into the machine now. I have done it before, during debug, but cant remember the credentials I used.

george1421

@adukes40 From the FOS engine console, set root’s password with passwd. Once that is set you should be able to putty in as root and what ever password you set.

Yeah the normal debug doesn’t pass the required parameters you need. The deploy debug will drop you at a linux prompt. Then key in fog<cr> and it will single step you through deployment. This what you want because you want to blow out of the deployment script where your post install script will run. This will give you a chance to check the drive geometry and actually key in (copy paste) your script until you reach the spot where it fails.

adukes40

@george1421 AH, so i have to set the password from the machine physically. That will have to wait i guess until Monday, don’t think I will be in at all this weekend.

george1421

@adukes40 Yeah, security its such a PITA some times.

This is the design, you don’t want someone hacking into the image deployment process, really. So this IS a security measure, but also gives the IT support people a way to debug their process, under their control and not by some common back door.

adukes40

@george1421 Oh I’m perfectly fine with the design. I just remember I puttyd in before on a debug machine…but could not remember how I did it.

Quazz

Is this the exact same image you’re trying to deploy?

An important thing to note is that if you have UEFI images (with GPT layout ofc), then you need to mount sda4 instead of sda2

adukes40

@Quazz We only have 3 images, and currently we are deploying the 32bit one. All the images are the same. plus this image in particular, is the one that was captured, as it is on the subnet where the master node resides.

Tomorrow i will be back at the building, so I will be able to dive into it more.

bigjim

I am having the same issue the pc has a 500GB drive so shouldnt the mount point show that as available? Am i missing a step?

Quazz

@adukes40 Any update on this?

@bigjim What model of computer? Model of hard drive? Are you also using a postdownloadscript for driver install? What exactly fails to mount?

adukes40

@Quazz I do not have anything further. Here is what I do have. this first script works for all devices with mechanical SATA, and SSD’s, but not M.2 SATA:

#!/bin/sh
osdiskpart=“/dev/sda2”;
driverver=“Win7”
mkdir /ntfs 2>/dev/null
mount.ntfs-3g “${osdiskpart}” /ntfs 2>/tmp/mntfail
mkdir /ntfs/Drivers 2>/dev/null
if [ -d “/ntfs/Windows/SysWOW64” ]
then
setarch=“x64”;
else
setarch=“x86”;
fi
machine=dmidecode -s system-product-name;
machine=“${machine%”${machine##[![:space:]]}“}”;
echo "Detected [${machine}] [${driverver}] with this arch [${setarch}] " >> /ntfs/Drivers/machine.txt
rm -f /tmp/mydrivers;
ln -s “/images/Drivers/${driverver}/${machine}/${setarch}/” /tmp/mydrivers;
if [ -d “/tmp/mydrivers” ]
then
cp -r /tmp/mydrivers/ /ntfs/Drivers;
fi
regfile=“/ntfs/Windows/System32/config/SOFTWARE”
key=“\Microsoft\Windows\CurrentVersion\DevicePath”
devpath=“%SystemRoot%\inf;C:\Drivers”;
reged -e “$regfile” &>/dev/null <<EOFREG
ed $key
$devpath
q
y
EOFREG
rm -f /tmp/mydrivers;
umount /ntfs

The following scirpt works with the M.2 SATA and mechanical SATAs, and the SSD’s. HOWEVER, it will not work with the OptiPlex 790’s nor 990’s They are the only two models giving the “space” problem.

#!/bin/sh
if [[ $hd == /dev/sda* ]]
then
osdiskpart=“/dev/sda2”;
else [[ $hd == /dev/nvme* ]]

osdiskpart="/dev/nvme0n1p2";

fi
driverver=“Win7”
mkdir /ntfs 2>/dev/null
mount.ntfs-3g “${osdiskpart}” /ntfs 2>/tmp/mntfail
mkdir /ntfs/Drivers 2>/dev/null
if [ -d “/ntfs/Windows/SysWOW64” ]
then
setarch=“x64”;
else
setarch=“x86”;
fi
machine=dmidecode -s system-product-name;
machine=“${machine%”${machine##[![:space:]]}“}”;
echo "Detected [${machine}] [${driverver}] with this arch [${setarch}] " >> /ntfs/Drivers/machine.txt
rm -f /tmp/mydrivers;
ln -s “/images/Drivers/${driverver}/${machine}/${setarch}/” /tmp/mydrivers;
if [ -d “/tmp/mydrivers” ]
then
cp -r /tmp/mydrivers/ /ntfs/Drivers;
fi
regfile=“/ntfs/Windows/System32/config/SOFTWARE”
key=“\Microsoft\Windows\CurrentVersion\DevicePath”
devpath=“%SystemRoot%\inf;C:\Drivers”;
reged -e “$regfile” &>/dev/null <<EOFREG
ed $key
$devpath
q
y
EOFREG
rm -f /tmp/mydrivers;
umount /ntfs

for now I reverted back to the first script for imaging as we only have 1 lab of the M.2 SATA machines. I have not had the time to look at this yet. If anyone has an idea on what to change for the scirpts I would be all ears, but for now I do not have a resolution. (due to time)

Quazz

@adukes40 said in No space left on device:

if [[ $hd == /dev/sda* ]]
then
osdiskpart=“/dev/sda2”;
else [[ $hd == /dev/nvme* ]]
osdiskpart=“/dev/nvme0n1p2”;

fi

I’m not an expert in bash scripts but this seems incorrect to me.

if [ $hd == /dev/sda* ]
then
osdiskpart="/dev/sda2"
else [ $hd == /dev/nvme* ]
osdiskpart="/dev/nvme0n1p2"
fi

Seems better already. They key part seems to be that the incorrect partition is selected on those Optiplexes. So perhaps the $hd thingy isn’t super reliable?

I would do something like :

if [ -b /dev/sda2 ]
then
osdiskpart="/dev/sda2"
else [ -b /dev/nvme0n1p2 ]
osdiskpart="/dev/nvme0n1p2"
else
echo "No usable partition detected!";
fi

Seeing as you’ll be using those partitions anyway, you might as well test against their existence in a direct fashion anyway, since it won’t be able to do anything if they don’t exist.

You only need ; in specific circumstances as well if I’m not mistaken, shouldn’t be necessary for simply setting variables.

george1421

@Quazz A couple of things come to mind here. First the script structure looks wrong (maybe I’m a class-A nit picker).

I would expect a bash script to look something similar

if [ $hd == /dev/sda* ]; then
  osdiskpart="/dev/sda2";
elif [ $hd == /dev/nvme* ];
  osdiskpart="/dev/nvme0n1p2";
fi

From there I’m not sure if wild cards are supported in the test as well as the if comparison should be against a string not /dev/sda*

With that said, I might rewrte that code as:

if [ $hd == *"/dev/nvme"* ]; then
  osdiskpart= "${hd}0n1p2";
else
  osdiskpart= "${hd}2";
fi

george1421

@george1421 But what I would suggest that you do when working on post download scripts is this.

Setup a debug deploy
PXE boot the target computer
After a few enter presses it will drop you to a command prompt
Give root a password with passwd
Get the IP address of the target computer with ip addr show
Go to your windows computer and open a putty session to your target computer using root and the password you just defined.

Now you can begin debugging your post install script.

From putty key in fog to start the download process.
Just past the point where the image has been sent to the target computer hit ctrl-C to break out of the fog installer script
Now the /images share should be mounted and you can test commands to find out the details of the hardware. By using putty you can copy and paste text easy.

adukes40

@george1421 Wish I had the time right now to run the tests, but with schools starting n a couple weeks, I might get to it by October. I can test your rewritten section though. I can pop that in and test a machine in a few seconds. It is just strange that 2 models, and only 2 models are being a pain. 745, 755, 760, and 780’s go thru without a hitch. BLEH!

cerebron

Maybe it’s not related to the topic, but a had an issue with free disk space after imaging from golden image. After “unsysprep” finished and i’ve logged on to machine, win7 shows that system partition is normaly extended to real hard disk size, but no matter how big hdd is, there is only 100-1000mb free on disk. After bashing my head for almost one week, i’ve realised that when i’ve created virtual disk on my ESXI virtual machine, for creating golden image, i’ve selected “Think provisioning, Lazy zeroed”, and so virtual disk file grows dynamicaly. I think because of this, part-clone cannot correctly detect partitions boundaries. So, the right choise is to select “Thick provisioning, Eager zeroed”.
Maybe that will help someone.

UPDATE:
Actualy, i’ve even figured out how to fix imaged machines disk size. Although, i dont know how to mass fix that, on imaged machine, you can first shrink system partition with acronis, for example, then extend it again.

Wayne Workman

@cerebron That is a great tip! Thank you for posting!

#wiki worthy

No space left on device

245

11.6k

17.1k

154.6k