10

I understand that solving long boot times involves analyzing how long it takes to boot up what, but the output of systemd-analyze blame and systemd-analyze plot has left me puzzled.

~ $ systemd-analyze
Startup finished in 12.557s (firmware) + 4.516s (loader) + 3.732s (kernel) + 26.720s (userspace) = 47.526s
~ $ systemd-analyze blame | grep "\s[1-9]*\."
          8.989s keyboard-setup.service
          8.757s dev-sda2.device
          6.055s apparmor.service
          4.948s accounts-daemon.service
          4.446s NetworkManager.service
          3.383s gpu-manager.service
          3.134s systemd-udevd.service
          3.079s snapd.firstboot.service
          2.440s udisks2.service
          2.249s grub-common.service
          2.093s upower.service
          1.943s networking.service
          1.661s avahi-daemon.service
          1.461s rsyslog.service
          1.460s pppd-dns.service
          1.449s systemd-tmpfiles-setup-dev.service
          1.387s systemd-rfkill.service
          1.290s colord.service
          1.210s resolvconf.service
          1.192s apport.service
          1.188s systemd-modules-load.service
          1.187s systemd-remount-fs.service
          1.166s dev-mqueue.mount
          1.152s bluetooth.service
          1.032s lightdm.service
          1.013s plymouth-quit-wait.service

Output of systemd-analyze plot

Information

The machine is a Dell Inspiron 5559; I've had it since February/March 2016.

~ $ uname -imporvs
Linux 4.8.0-32-generic #34-Ubuntu SMP Tue Dec 13 14:30:43 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

Distro is Lubuntu 16.10 w/LXDE.

~ $ sudo parted /dev/sda unit mib print
Model: ATA ST1000LM024 HN-M (scsi)
Disk /dev/sda: 953870MiB
Sector size (logical/physical): 512B/4096B
Partition Table: gpt
Disk Flags: 

Number  Start      End        Size       File system     Name                  Flags
 1      1.00MiB    513MiB     512MiB     fat32           EFI System Partition  boot, esp
 2      513MiB     937591MiB  937078MiB  ext4
 3      937591MiB  953869MiB  16278MiB   linux-swap(v1)

Worst part is, the times of the individual modules vary a bit (1 to 2 seconds, observed from following this problem since I installed Lubuntu), which means I would need to update systemd-analyze blame constantly or log a series of reboots and then make an average.

Can anyone tell me where I could start?

UPDATE

Upgrading from 16.10 to 17.04 via sudo apt dist-upgrade changed the situation considerably.

~ $ systemd-analyze blame | grep "\s[1-9]*\."
         16.083s dev-sda2.device
         15.435s keyboard-setup.service
          8.015s systemd-udevd.service
          4.090s NetworkManager.service
          3.644s systemd-tmpfiles-setup-dev.service
          2.621s apparmor.service
          2.549s grub-common.service
          2.477s plymouth-read-write.service
          1.560s accounts-daemon.service
          1.107s systemd-modules-load.service
          1.002s colord.service
~ $ systemd-analyze critical-chain
The time after the unit is active or started is printed after the "@" character.
The time the unit takes to start is printed after the "+" character.

graphical.target @25.631s
└─multi-user.target @25.631s
  └─getty.target @25.631s
    └─getty@tty1.service @25.631s
      └─system-getty.slice @25.630s
        └─setvtrgb.service @25.407s +222ms
          └─systemd-user-sessions.service @25.245s +2ms
            └─network.target @25.245s
              └─NetworkManager.service @21.154s +4.090s
                └─dbus.service @21.147s
                  └─basic.target @21.139s
                    └─sockets.target @21.139s
                      └─snapd.socket @21.136s +2ms
                        └─sysinit.target @21.110s
                          └─apparmor.service @18.488s +2.621s
                            └─local-fs.target @18.488s
                              └─boot-efi.mount @18.387s +100ms
                                └─systemd-fsck@dev-disk-by\x2duuid-7930\x2d6EDD.service @18.198s +150ms
                                  └─dev-disk-by\x2duuid-7930\x2d6EDD.device @18.198s

Output of systemd-analyze plot At least clear culprits are appearing.

CLOSED

The post is being closed because I have migrated to another distro (Gentoo) where the problem has not arisen, so the question is no longer relevant.

  • Ok, one lead I have is that some of the services mentioned by systemd-analyze blame (in particular keyboard-setup.service) are SysVInit-style scripts located in /etc/init.d. Though I don't know how you would replace a script-based service... – setun-90 Jan 3 '17 at 9:33
  • grep "\s[1-9]\." any reason you're filtering out services with >10s load times? Put a + after the ] to match one or more digits. – Jacob Krall Jan 3 '17 at 12:06
  • @JacobKrall I didn't exactly filter them out, it's just I didn't have any services with > 10s load times, hence the single digit. I did this in a hurry... and '+' didn't work for me, '*' did. – setun-90 Jan 3 '17 at 12:14
  • Okay, sorry for the bother. That's strange that + didn't work; it's one of the repetition operators in GNU Grep gnu.org/software/grep/manual/grep.html#Fundamental-Structure – Jacob Krall Jan 3 '17 at 12:25
  • @JacobKrall I also thought it was strange too. Debug later. – setun-90 Jan 3 '17 at 12:27
1

Can anyone tell me where I could start?

Run a Live Ubuntu Session (or any distro that comes with "try without installing" feature)

Many a times Linux based distros take long time to boot or even fail to boot when there's some issue with a peripheral component like keyboard or NIC, etc. For example, my old laptop's keyboard's "Up" Key remains in pressed state without being physically pressed. Because of this the keyboard-setup.sh waits for a long time, fails to complete and finally I see a bunch of error messages that notify me of Ubuntu not being able to boot. Disconnecting keyboard during boot was the workaround for me to make it boot.

Testing your hardware for such type of errors would be a good starting point. If you know about a hardware issue with your laptop you can try to disconnect that component during boot (probably NIC or keyboard because you mentioned polktid and keyboard-setup.sh )

  • Thanks for mentioning the hardware, I hadn't thought of that. Though I should also have mentioned in the question that I did a distro upgrade to 17.04 and the boot times have changed slightly (with udevd now being the main culprit), but I do think keyboard-setup.sh is still taking a long time. I'll update. – setun-90 May 16 '17 at 6:49
  • Pls mention that in your question. Which version you upgraded from? Upgrading from LTS to a Release always causes problems. If you upgraded from 16.xx LTS to 17.04 you will have to do a clean install of 17.04 then. I insist to try a live session of 17.04. If live session is booting fine, a clean install will definitely fix things up. – sziraqui May 17 '17 at 13:57
  • Sorry, I did the upgrade in the meantime, after this question was asked. Boot times actually shortened by a second or two. But yes, I guess a clean reinstallation could do something. And btw I thought 16.10 wasn't LTS. – setun-90 May 18 '17 at 14:19
  • Another point to note, you cannot officially update from an LTS (eg 16.xx, 14.xx) to a Release (eg 15.xx, 17.xx) or vice versa. You can update with an iso ofcourse but it always makes the system buggy. I guessed you have upgraded from iso and that's why I suggested to do a clean install. If this is the case, I will update my answer which might help some one else in future. – sziraqui May 19 '17 at 15:25
  • I didn't use an ISO, the offer to upgrade showed up one day through Synaptic and I then ran sudo apt dist-upgrade. – setun-90 May 20 '17 at 16:14

Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.