1

I am running a server with Debian 8, and after running modprobe nbd, the program runs just fine. However, the device is not created since /dev/nbd0 does not exist. What should I do to get modprobe to run properly?

SalmonKiller
  • 113
  • 1
  • 5
  • I have a feeling that you're comparing an old version to the system you're on. The fact that your application can access the networked drive is already proof enough that it is working. – Julie Pelletier Jul 02 '16 at 17:39

1 Answers1

4

I have a partial answer that effectively works around your problem but does not conclusively explain why the problem happened in the first place.

Resolution

Run these commands:

sudo rmmod nbd 
sudo mount -t devtmpfs none /dev 
sudo modprobe nbd 
ls /dev/nbd*

The final command should look like this:

root@node51 [~]# ls /dev/nbd*
/dev/nbd0  /dev/nbd1  /dev/nbd10  /dev/nbd11  /dev/nbd12  /dev/nbd13  /dev/nbd14  /dev/nbd15  /dev/nbd2  /dev/nbd3  /dev/nbd4  /dev/nbd5  /dev/nbd6  /dev/nbd7  /dev/nbd8  /dev/nbd9

Explanation

We determined in chat that /dev was not being updated because it was mounted as tmpfs instead of as devtmpfs.

You can check the second column of the following command to see if /dev is mounted as tmpfs or devtmpfs:

df -T /dev

Without devtmpfs, only the device and character files defined during the initial boot would be populated in /dev. devtmpfs allows devices to be added and removed after boot.

It's not possible to unmount /dev with umount /dev because the special files in there are in use, but it is possible to load a fresh devtmpfs over the existing /dev mount.

sudo mount -t devtmpfs none /dev mounts a devtmpfs over the existing mount at /dev.

Now, when you do sudo modprobe nbd, the device files /dev/nbd0 through /dev/nbd15 get populated in /dev.

Cause

I did not determine why /dev was mounted as tmpfs instead of devtmpfs or why udev wasn't running. devtmpfs should have been the default.

This answer on Server Fault might have some leads.

Community
  • 1
Deltik
  • 314
  • 1
  • 4
  • 14
  • Hmm. I tried doing this, and then doing `sudo rmmod nbd`. It actually removed the nbd from somewhere (I don't know where). Given that i'm running the whole system on google cloud platform, any tips on where the nbd might be? – SalmonKiller Jul 02 '16 at 22:24
  • @SalmonKiller: Do you have the stock kernel that ships with Debian 8? Does `nbd: registered device at major` show up in `/var/log/syslog`? – Deltik Jul 02 '16 at 22:53
  • Nope, it doesn't show up in `/var/log/syslog`. – SalmonKiller Jul 02 '16 at 23:29
  • @SalmonKiller: And what kernel are you running? – Deltik Jul 03 '16 at 00:16
  • 3.16.0-4-amd64 is the kernel – SalmonKiller Jul 03 '16 at 00:31
  • @SalmonKiller: Assuming you're logging normally, the nbd kernel module is not reaching [this line in the source code](https://github.com/torvalds/linux/blob/04cfac4e40b2b5dbaf8c0fc625878388a98bb9c6/drivers/block/nbd.c#L862). You should get network block devices because every failure case above that line would return an error code. Do you have folders at `/sys/block/nbd*`? – Deltik Jul 03 '16 at 03:14
  • Yeah, i have those directories – SalmonKiller Jul 03 '16 at 05:10
  • Let us [continue this discussion in chat](http://chat.stackexchange.com/rooms/41966/discussion-between-deltik-and-salmonkiller). – Deltik Jul 03 '16 at 05:23
  • i had the same problem in Ubuntu 20.04, but for me it was sufficient to just run `sudo rmmod nbd; sudo modprobe nbd;` thanks – hanshenrik Jan 13 '21 at 23:10