Ani's questions

I have been using df -lh -x nfs --total to get the total disk available (with file-systems) and it has served me well, so far. With systemd, I see that there are entries in df for each logged in user:

$ df -lh -x nfs --total
Filesystem               Size  Used Avail Use% Mounted on
udev                     3.7G     0  3.7G   0% /dev
tmpfs                    758M   11M  748M   2% /run
/dev/mapper/system-root   78G  8.6G   65G  12% /
tmpfs                    3.8G  4.0K  3.8G   1% /dev/shm
tmpfs                    5.0M     0  5.0M   0% /run/lock
tmpfs                    3.8G     0  3.8G   0% /sys/fs/cgroup
tmpfs                     16M     0   16M   0% /mnt/my_ramdisk
tmpfs                    758M     0  758M   0% /run/user/1003     <<<<
tmpfs                    758M     0  758M   0% /run/user/1001     <<<<
total                     91G  8.6G   79G  10% -             

So, the total is off!

Should I just exclude the tmpfs with -x tmpfs or is there any better way to only exclude the systemd generated mounts ?

Thanks.

I have read the systemd man pages and also some Q&A here related to StartLimit*

I tried what is recommended for newer versions:

[Unit]
StartLimitBurst=5
StartLimitIntervalSec=90

[Service]
RestartSec=5
Restart=always

Also, what is supposed to work for older versions of systemd:

[Service]
StartLimitBurst=5
StartLimitInterval=90
...
RestartSec=5
Restart=always

(this is still accepted and systemd-analyze does not show any errors)

But neither of this works! i.e. my daemon keeps indefinitely restarting I can see NRestarts also showing the correct count.

I am on version systemd 237 on Ubuntu 18.04 (Bionic). I don't mind using either format, but I want to make sure that the restart attempt is capped to 5 or 6 times and no more.

I am not seeing any messages on bootup/shutdown on Bionic (18.04). Do I have to add systemd.show_status=yes on the kernel command line or just the config file /etc/systemd/system.conf's ShowStatus=yes change - sufficient ?

On Ubuntu Bionic (18.04) I do not see any messages during boot-up or shutdown on the text console! (virsh console). However, I do see messages on the graphical console (virt-manager GUI)

Do I need console=tty0 in the kernel command line ? I have [email protected] enabled and running.

I have a need to install customized kernel package set (the usual linux-modules-<ver>.deb, linux-modules-extra-<ver>.deb, linux-headers-<ver>.deb, linux-image-<ver>.deb...). I install it on Ubuntu Bionic (18.04.6). It installs and things look normal, but once I reboot the VM, I see that the packages from upstream would have got installed (overwriting mine)!! I see in /var/log/dpkg.log, logs like:

startup archives unpack

...

startup packages configure

I thought that this could have been cached and tried to clear /var/cache/apt/archives, but still the packages are overridden! Also, before a reboot, when I checked all the packages were in ii state, so its not like any of them were in bad state and needed a re-install!

I am not sure why this is happening! Can anyone throw some light on this startup package install please? I just want to make sure that my customized packages override the ones from upstream.

I am running Ubuntu Bionic as a VM on KVM, and after a few days, the VM suddenly becomes slow, when I run top, I see dbus-daemon (high CPU) and polkitd.

A reboot fixes it, but I want to know if there is something that I can do to "fix" the host without a reboot.

I have a host where I have installed apt-cacher-ng and copied over a pre-populated cache under /var/cache/apt-cacher-ng. I see that the service is running with no errors. However, when I try to download something, apt still seems to try and use archive.ubuntu.com ! This host does not have network connectivity. How can I make sure to tell apt to use only the local cache ?

I have also added the following config:

$ cat /etc/apt/apt.conf.d/99_local_proxy.conf
Acquire::http::Proxy "http://localhost:3142/";
Acquire::ForceIPv4 "true";
Acquire::Retries "3";

This is the error I get:

W: Failed to fetch http://us.archive.ubuntu.com/ubuntu/dists/bionic/InRelease  503  DNS error for hostname us.archive.ubuntu.com: Temporary failure in name resolution. If us.archive.ubuntu.com refers to a configured cache repository, please check the corresponding configuration file. [IP: 127.0.0.1 3142]

I have a service running on Bionic (18.04.5), which requires that /mnt/huge be mounted. This service runs most often, but in some rare cases it fails and I see that the hugetlbfs is not mounted on the system. Do I have to add a line like

Requires=mnt-huge.mount

To my unit file ? I see that there is already:

hugetlbfs /mnt/huge hugetlbfs defaults 0 0

in /etc/fstab

As per this article: https://www.linux-kvm.org/page/USB I can create a disk image and attach it as a USB drive to my running VM.

I tried this, via qemu-monitor-command option of virsh, but somehow wherever I place the img file, its unable to access!

$ sudo  virsh qemu-monitor-command --hmp MyVM1 'drive_add 0 id=my_usb_disk,if=none,file=/tmp/usb.img'
Could not open '/tmp/usb.img': Permission denied
$ ls -l /tmp/usb.img 
-rwxrwxrwx 1 libvirt-qemu kvm 8388608000 Apr 16 19:02 /tmp/usb.img

Unless I run this, I dont think I can do device_add ?!

This is my qemu version (on Bionic - 18.04):

$ qemu-system-x86_64 --version 
QEMU emulator version 2.11.1(Debian 1:2.11+dfsg-1ubuntu7.4)

Is there any other way (say, libvirt XML) where I can attach a raw image as a USB drive? (I searched in libvirt docs, I could not find any example for this)

What am I missing here ?

-- Thanks

With help from my other Q&A here, I am able to get the ISO to boot, but not install it (yet :-/).

As per some Launchpad posts, the normal initrd present in the ISO will always look for CD-ROM. So, I replaced it with the initrd designed for hd-media from here:

http://archive.ubuntu.com/ubuntu/dists/bionic/main/installer-amd64/current/images/hd-media/initrd.gz http://archive.ubuntu.com/ubuntu/dists/bionic/main/installer-amd64/current/images/hd-media/vmlinuz

(although, I am not sure what is the exact difference) With this, I could get past the CD-ROM mount, but installation gets into mount problem. My ISO is on /dev/sda5, so when asked to unmount /dev/sda, I said "No", later it goes a step further and errors out like "no root file system defined" etc.

Tried: https://bugs.launchpad.net/ubuntu/+source/debian-installer/+bug/1347726 Didn't help. Also tried using toram in the command line as per this - didn't help.

I am using Ubuntu 18.04.4 server ISO here. However, just to try, I used Ubuntu 18.04.5 Desktop (live CD boot) and that installed fine (even though it gave the error for mounted /dev/sda!) What exactly is the delta that is missing in Server ISO that is present in Desktop /Live ISO which makes this work!?

All I want is to install an ISO which is residing on one partition /dev/sda5 to another - /dev/sda2 or /dev/sda6

Please Note: I read all posts in here, about booting ISO from partition using GRUB2 and based on that, I have tried the following, which is not worknig:

I have these partitions and I want to boot an ISO from /dev/sda5 -- should be (hda0,5) in GRUB terms.

root@myhost:/home/vyom# lsblk 
NAME   MAJ:MIN RM   SIZE RO TYPE MOUNTPOINT
sda      8:0    0    30G  0 disk 
`-sda1   8:1    0   243M  0 part [SWAP]
`-sda2   8:2    0  11.2G  0 part /
`-sda3   8:3    0     1K  0 part 
`-sda5   8:5    0   9.3G  0 part /mypart5    <<<
`-sda6   8:6    0   9.3G  0 part /mypart6
root@myhost:/home/vyom# blkid
/dev/sda1: UUID="79813f4a-85da-4c81-a044-097922b30648" TYPE="swap" 
/dev/sda2: UUID="68923f34-385e-4740-b8af-4502aa3dd847" TYPE="ext4" 
/dev/sda5: UUID="ef7ca072-bfc9-457b-8a9e-b923fa0d3fe7" TYPE="ext4"   <<<
/dev/sda6: UUID="d8bd9333-dc41-47b3-8a42-f6c308a6f047" TYPE="ext4" 

I tried using UUID with --fs-uuid and fromiso= - didn't work:

menuentry 'UbuntuBionicVSCustom4' --class ubuntu --class gnu-linux --class gnu --class os {
   set isofile="/mypart5/ubuntu-18.04-3f8806d-test2.iso" 
   insmod ext2
   insmod loopback
   insmod iso9660
   loopback loop (hd0,5)$isofile
   search --no-floppy --fs-uuid --set=root ef7ca072-bfc9-457b-8a9e-b923fa0d3fe7
   linux (loop)/install/vmlinuz boot=install fromiso=/dev/sda5/$isofile noprompt noeject
   initrd (loop)/install/initrd.gz
}

I tried --file option to search - that too didnt work:

menuentry 'UbuntuBionicVSCustom3' --class ubuntu --class gnu-linux --class gnu --class os {
   set isofile="/mypart5/versa-flexvnf-3f8806d-21.2.1-B-S.iso"
   insmod ext2
   insmod part_msdos
   insmod lvm
   insmod loopback
   insmod iso9660
   search --file --no-floppy --set=root "${isofile}"
   loopback loop "(${root})${isofile}"
   linux (loop)/install/vmlinuz boot=install findiso="${isofile}" noprompt noeject
   initrd (loop)/install/initrd.gz
}

The ISO file is actually at: /mypart5/ubuntu-18.04-3f8806d-test2.iso

So, should I do:

   set isofile="/mypart5/ubuntu-18.04-3f8806d-test2.iso"

OR, just:

   set isofile="ubuntu-18.04-3f8806d-test2.iso"

So that, next I can do:

   loopback loop (hd0,5)$isofile

also, I am confused with the kernel command line options, should I use:

     fromiso=/dev/sda5/$isofile

OR

     findiso="${isofile}"

OR

     iso-scan/filename=${isofile}

OR

     isoloop=/${isofile}

I am not able to get GRUB to actually pick the file from the partition /dev/sda5 mounted on /mypart5. I get "error: no such device" or unknown filesystem . Please help me what I am missing here!

I am on Bionic (18.04.4) and I see that my jobs are stuck in the queue and never run!

[root@myhost2: log]# atq
3267   Tue Jun 16 03:34:00 2020 = root
697    Mon Jun 15 19:36:00 2020 = root
3268   Tue Jun 16 03:34:00 2020 = root
3265   Tue Jun 16 03:31:00 2020 = root
7566   Tue Jun 16 22:17:00 2020 a root
7563   Tue Jun 16 22:16:00 2020 a root
697    Mon Jun 15 19:36:00 2020 a root

[root@myhost2: log]# service atd status 
* atd.service - Deferred execution scheduler
   Loaded: loaded (/lib/systemd/system/atd.service; enabled; vendor preset: enabled)
   Active: active (running) since Tue 2020-06-16 14:03:35 PDT; 8h ago
     Docs: man:atd(8)
Main PID: 731 (atd)
    Tasks: 1 (limit: 4915)
   CGroup: /system.slice/atd.service
       `-731 /usr/sbin/atd -f

Jun 16 22:20:21 myhost2 atd[19018]: File a002b90194efbc is in wrong format - aborting
Jun 16 22:20:21 myhost2 atd[19019]: File a002b90194efbc is in wrong format - aborting
Jun 16 22:20:21 myhost2 atd[19020]: File a002b90194efbc is in wrong format - aborting
Jun 16 22:20:21 myhost2 atd[19021]: File a002b90194efbc is in wrong format - aborting

The file being pointed to (a002b90194efbc) is empty!

If I delete the files in spool and reboot it becomes ok, but not sure when it might go back to the same bad state!

How do I make sure to never get into this stuck state!? Or, if stuck, any way to auto recover?

-- Thanks

I am installing a Ubuntu ISO (from USB [burnt with Rufus]), which installs fine but grub is installed on /dev/sda instead of sdb, so the system becomes like live boot - have to always keep USB plugged in! If I run grub-install /dev/sdb after installation, then from next reboot it works fine! Is there anyway to tell grub to install on dev/sdb or disk ?

P.S. the device has UEFI, but I am using Legacy bios if that matters.

I am not able to figure out what option to pass in StandardOutput= for a unit file (.service), where I want to show some messages on the connected terminal from where the service is started! (console/tty doesn't seem to be what I want)

Maybe /etc/systemd/system.conf's global default DefaultStandardOutput=journal can be changed, but I don't want to do that! I am only interested in showing some progress/startup messages from my unit file (like systemd-run -P)

I am on Ubuntu 18.04.2

# systemd --version
systemd 237

I am migrating our software from Trusty to Bionic. On trusty, there were many dependencies which were manually installed, but many of those were removed/renamed in Xenial itself and not available on Bionic. For e.g., gdb-minimal, python-support etc. Is there any easy way to find the equivalent packages for these, on Bionic ? Or, will some of them have to built for Bionic if they are mandatory ?

I want to be able to set timezone for Ubuntu hosts, from command line (given in POSIX format). I was searching for options, and found timedatectl from systemd package but since I did not want the entire systemd, but just timedatectl, I downloaded the .deb package and extracted the timedatectl binary, but that expects GLIBC>=2.17, I have GLIBC 2.15 on my Ubuntu 12.04 !

Is there any other utility, which is simple enough to use, that I can install, to set timezone in POSIX format (e.g. EST+10)

Upstart has job environment and event environment, I am faimilar with the jobs, and I can write script stanzas to log/exec. But when the process fails, there is a stopped event, which has some useful environment variables - like EXIT_STATUS/EXIT_SIGNAL, which I don't know if I can log!. These are not available in the post-stop/pre-stop stanzas. So, is there any other way where I can access these variables and log them?

Thanks.

I have a customized Ubuntu ISO, where I have all my required packages being installed, from the /pool/extras/ directory. Now, I want to be able to copy one of the packages (.deb) from this /pool/extras directory and save it to another location, say to the home directory of a user that I know exists (/home/myuser/). I tried adding this in the preseed

d-i   preseed/late_command string if [ -f /media/cdrom/pool/extras/mypkg.deb ]; then cp -f /media/cdrom/pool/extras/mypkg.deb /home/myuser; fi

But this didn't seem to work!, is this approach correct ? or is there any other way that I can get the package (at a later stage, maybe) from /pool/extras ?

Is it possible to have an interface definition for say eth0 in /etc/network/interfaces.d/eth0.cfg and a section in /etc/network/interfaces?

If this is possible, is there a precedence for one over the other?