How can I decrease my boot time? (60+ sec)

I-sty · 2014-02-20 15:22:59

Feb 20 17:12:30 laptop systemd[1]: Starting CUPS Printing Service Sockets.
Feb 20 17:12:30 laptop systemd[1]: Listening on CUPS Printing Service Sockets.

Here I lose 30 seconds. :(

Feb 20 17:12:58 laptop systemd[1]: Starting D-Bus System Message Bus Socket.
Feb 20 17:12:58 laptop systemd[1]: Listening on D-Bus System Message Bus Socket.

[i-sty@laptop ~]$ systemd-analyze blame && systemd-analyze time && systemd-analyze critical-chain
          1.621s udisks2.service
          1.523s systemd-vconsole-setup.service
          1.111s run-media-i\x2dsty-Egyeb.mount
           868ms systemd-udev-trigger.service
           857ms systemd-tmpfiles-setup.service
           853ms kmod-static-nodes.service
           851ms dev-mqueue.mount
           851ms dev-hugepages.mount
           610ms systemd-readahead-replay.service
           610ms systemd-readahead-collect.service
           609ms sys-kernel-debug.mount
           477ms lightdm.service
           403ms colord.service
           383ms systemd-logind.service
           330ms dhcpcd.service
           285ms systemd-tmpfiles-setup-dev.service
           237ms sys-kernel-config.mount
           216ms systemd-sysctl.service
           193ms user@1000.service
           126ms systemd-udevd.service
           117ms systemd-journal-flush.service
            44ms systemd-random-seed.service
            30ms systemd-update-utmp.service
            28ms systemd-readahead-done.service
            20ms var-tmp.mount
            19ms tmp.mount
            13ms polkit.service
             8ms alsa-restore.service
             4ms systemd-user-sessions.service
             4ms upower.service
             2ms rtkit-daemon.service
             2ms bluetooth.service
             1ms systemd-remount-fs.service
             1ms sys-fs-fuse-connections.mount
Startup finished in 4.819s (kernel) + 31.353s (userspace) = 36.172s
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 @31.352s
└─lightdm.service @30.875s +477ms
  └─systemd-user-sessions.service @30.870s +4ms
    └─basic.target @30.865s
      └─timers.target @30.864s
        └─systemd-tmpfiles-clean.timer @30.863s
          └─sysinit.target @2.958s
            └─systemd-update-utmp.service @2.927s +30ms
              └─systemd-tmpfiles-setup.service @2.070s +857ms
                └─local-fs.target @2.069s
                  └─var-tmp.mount @2.049s +20ms
                    └─local-fs-pre.target @1.923s
                      └─systemd-tmpfiles-setup-dev.service @1.638s +285ms
                        └─kmod-static-nodes.service @784ms +853ms
                          └─systemd-journald.socket @767ms
                            └─-.mount @765ms
                              └─system.slice @1.376s
                                └─-.slice @1.375s

My fstab file:

# /etc/fstab: static file system information.
#
# Use 'blkid' to print the universally unique identifier for a
# device; this may be used with UUID= as a more robust way to name devices
# that works even if disks are added and removed. See fstab(5).
#
# <file system> <mount point>   <type>  <options>       <dump>  <pass>
#
#UUID=e4c124f1-b01a-4f2c-aede-cfc6d1860bbf swap swap defaults 0 0
#UUID=2d84dcc8-042a-4056-bde2-720b278877aa / btrfs rw,relatime 0 0

/dev/sda6         /run/media/i-sty/Egyeb     ntfs-3g auto,uid=i-sty,gid=users,relatime,noauto,x-systemd.automount     0 0
UUID=E42AB6452AB61490    /run/media/i-sty/Filmek ntfs-3g auto,uid=i-sty,gid=users,relatime,noauto,x-systemd.automount     0 0
tmpfs           /tmp                 tmpfs   defaults,noatime,nodev,nosuid,size=3G                0 0
tmpfs            /var/tmp        tmpfs    defaults,noatime,size=2G                    0 0

Systemd/ctl shows ~40 seconds to boot time, but in fact is >60seconds.
How can improve my system?

Last edited by I-sty (2014-02-20 15:23:39)

karol · 2014-02-20 15:28:00

It's not bad for a laptop hard drive.

I-sty · 2014-02-20 15:33:09

karol wrote:

It's not bad for a laptop hard drive.

>60 seconds not bad?
Okay, I don't want 5 seconds, I have just a 5400rpm WD hard not an SSD;
But why I lose 30 seconds in the userspace?

tydell · 2014-02-20 16:05:17

I-sty wrote:

karol wrote:
It's not bad for a laptop hard drive.

>60 seconds not bad?
Okay, I don't want 5 seconds, I have just a 5400rpm WD hard not an SSD;
But why I lose 30 seconds in the userspace?

I must say that I had 8 seconds boot time on my old Acer 3690 WLMI with 64GB 5400rpm hd. So >60 seconds is terrible time.

I-sty · 2014-02-20 16:19:23

http://imagerz.com/QBZEDktvAwMCA1NMRwVQ

http://imagerz.com/QBZEDktvAwMCA1NPEQVQ

-- mod edit: read the Forum Etiquette and only post thumbnails http://wiki.archlinux.org/index.php/For … s_and_Code [jwr] --

karol · 2014-02-20 16:32:21

Please use thumbnails rather than big screenshots: https://wiki.archlinux.org/index.php/Fo … s_and_Code

drcouzelis · 2014-02-20 17:02:35

I-sty, do you have "cups.service" or "cups.socket" enabled through systemd?

If you (temporarily) disable them, does your computer have an acceptable boot time?

I-sty · 2014-02-20 17:54:16

karol wrote:

Please use thumbnails rather than big screenshots: https://wiki.archlinux.org/index.php/Fo … s_and_Code

Sorry.

drcouzelis wrote:

I-sty, do you have "cups.service" or "cups.socket" enabled through systemd?
If you (temporarily) disable them, does your computer have an acceptable boot time?

Yes, I have cups. I disabled and rebooted my laptop few times but I had no difference. (65 ~ 85 sec)

          2.454s lightdm.service
          1.949s systemd-tmpfiles-setup.service
          1.530s udisks2.service
          1.397s systemd-logind.service
           929ms systemd-vconsole-setup.service
           703ms systemd-udev-trigger.service
           627ms systemd-readahead-replay.service
           627ms systemd-readahead-collect.service
           626ms sys-kernel-debug.mount
           592ms kmod-static-nodes.service
           591ms dev-mqueue.mount
           591ms dev-hugepages.mount
           411ms user@1000.service
           328ms upower.service
           305ms run-media-i\x2dsty-Egyeb.mount
           294ms dhcpcd.service
           200ms systemd-journal-flush.service
           172ms sys-kernel-config.mount
           143ms polkit.service
           137ms systemd-sysctl.service
           137ms systemd-remount-fs.service
           126ms systemd-tmpfiles-setup-dev.service
           122ms systemd-udevd.service
             5ms alsa-restore.service
             2ms bluetooth.service
             2ms systemd-random-seed.service
             2ms systemd-user-sessions.service
             2ms rtkit-daemon.service
             2ms tmp.mount
             1ms sys-fs-fuse-connections.mount
             1ms var-tmp.mount
             1ms systemd-readahead-done.service
             1ms systemd-update-utmp.service
Startup finished in 4.909s (kernel) + 38.309s (userspace) = 43.219s
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 @38.309s
└─lightdm.service @35.854s +2.454s
  └─systemd-user-sessions.service @35.850s +2ms
    └─basic.target @35.845s
      └─timers.target @35.844s
        └─systemd-tmpfiles-clean.timer @35.721s
          └─sysinit.target @3.783s
            └─systemd-update-utmp.service @3.781s +1ms
              └─systemd-tmpfiles-setup.service @1.831s +1.949s
                └─local-fs.target @1.831s
                  └─tmp.mount @1.828s +2ms
                    └─local-fs-pre.target @1.810s
                      └─systemd-tmpfiles-setup-dev.service @1.683s +126ms
                        └─kmod-static-nodes.service @1.090s +592ms
                          └─systemd-journald.socket @842ms
                            └─-.mount @831ms
                              └─system.slice @1.467s
                                └─-.slice @1.466s

Why the

└─systemd-tmpfiles-clean.timer @35.721s

is so long?
I use tmpfs to /tmp and /var/tmp

Edit:

Journalctl log here: http://pastebin.com/RwPEQmXf Look at the 740th line

Last edited by I-sty (2014-02-20 17:59:29)

Spider.007 · 2014-02-20 20:30:35

Yeah something weird is going on. Too bad you can't blame a single unit; did you have a look at https://wiki.archlinux.org/index.php/Bootchart ? It might just reveal what systemd is waiting for

I-sty · 2014-02-21 16:07:32

http://imagerz.com/QBZFWEtvAwMCA1MZQgVQ
Here is my bootchart file. I don't see any errors. (this file shows just 24 seconds).

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 @25.508s
└─lightdm.service @24.913s +594ms
  └─systemd-user-sessions.service @24.909s +3ms
    └─basic.target @24.899s
      └─timers.target @24.897s
        └─systemd-tmpfiles-clean.timer @24.897s
          └─sysinit.target @3.902s
            └─systemd-update-utmp.service @2.784s +1.117s
              └─systemd-tmpfiles-setup.service @2.275s +508ms
                └─local-fs.target @2.275s
                  └─var-tmp.mount @2.273s +2ms
                    └─local-fs-pre.target @2.257s
                      └─systemd-tmpfiles-setup-dev.service @1.950s +306ms
                        └─kmod-static-nodes.service @1.017s +932ms
                          └─systemd-journald.socket @964ms
                            └─-.mount @952ms
                              └─system.slice @1.616s
                                └─-.slice @1.616s

          1.284s systemd-vconsole-setup.service
          1.117s systemd-update-utmp.service
          1.055s sys-kernel-debug.mount
          1.023s upower.service
          1.003s dev-mqueue.mount
          1.002s dev-hugepages.mount
           932ms kmod-static-nodes.service
           711ms systemd-udev-trigger.service
           666ms udisks2.service
           654ms systemd-readahead-replay.service
           653ms systemd-readahead-collect.service
           600ms colord.service
           594ms lightdm.service
           508ms systemd-tmpfiles-setup.service
           390ms sys-kernel-config.mount
           339ms run-media-i\x2dsty-Egyeb.mount
           338ms systemd-sysctl.service
           308ms user@1000.service
           306ms systemd-tmpfiles-setup-dev.service
           294ms systemd-journal-flush.service
           266ms systemd-remount-fs.service
           240ms systemd-logind.service
           218ms dhcpcd.service
           182ms systemd-readahead-done.service
           180ms sys-fs-fuse-connections.mount
           133ms polkit.service
            73ms systemd-udevd.service
            19ms systemd-random-seed.service
            10ms alsa-restore.service
             3ms systemd-user-sessions.service
             2ms var-tmp.mount
             2ms bluetooth.service
             2ms rtkit-daemon.service
             1ms tmp.mount
Startup finished in 4.465s (kernel) + 25.508s (userspace) = 29.973s

Daw11 · 2014-02-21 18:07:46

I had a similar problem some time ago, I'm not sure if this is the correct solution to your problem but cleaning my journal log by deleting all files inside /var/log/journal/ and setting the max journal size in /etc/systemd/journald.conf at 50Mb with

SystemMaxUse=50M

speeded up my boot.
You can check your journal disk usage with

journalctl --disk-usage

edit: journalctl --disk-usage should be launched as root

Last edited by Daw11 (2014-02-21 22:17:02)

I-sty · 2014-02-21 18:18:55

Daw11 wrote:

I had a similar problem some time ago, I'm not sure if this is the correct solution to your problem but cleaning my journal log by deleting all files inside /var/log/journal/ and setting the max journal size in /etc/systemd/journald.conf at 50Mb with
SystemMaxUse=50M 
speeded up my boot.
You can check your journal disk usage with
journalctl --disk-usage

Thanks. I will try.

[i-sty@laptop ~]$ journalctl --disk-usage
Journals take up 423.6M on disk.

This is my journal conf.

[Journal]
Storage=persistent
Compress=yes
#Seal=yes
#SplitMode=login
#SyncIntervalSec=5m
RateLimitInterval=60s
#RateLimitBurst=1000
#SystemMaxUse=50M
#SystemKeepFree=
#SystemMaxFileSize=
#RuntimeMaxUse=
#RuntimeKeepFree=
#RuntimeMaxFileSize=
#MaxRetentionSec=
#MaxFileSec=1month
#ForwardToSyslog=yes
#ForwardToKMsg=no
#ForwardToConsole=no
#TTYPath=/dev/console
#MaxLevelStore=debug
#MaxLevelSyslog=debug
#MaxLevelKMsg=notice
#MaxLevelConsole=info

Last edited by I-sty (2014-02-21 18:21:14)

karol · 2014-02-21 18:48:28

Run ' journalctl --disk-usage' with sudo or as root.

I-sty · 2014-02-21 18:54:32

karol wrote:

Run ' journalctl --disk-usage' with sudo or as root.

[i-sty@laptop ~]$ sudo journalctl --disk-usage
Journals take up 2.5M on disk

Daw11 · 2014-02-21 19:17:49

I-sty wrote:

This is my journal conf.

[Journal]
Storage=persistent
Compress=yes
#Seal=yes
#SplitMode=login
#SyncIntervalSec=5m
RateLimitInterval=60s
#RateLimitBurst=1000
#SystemMaxUse=50M
#SystemKeepFree=
#SystemMaxFileSize=
#RuntimeMaxUse=
#RuntimeKeepFree=
#RuntimeMaxFileSize=
#MaxRetentionSec=
#MaxFileSec=1month
#ForwardToSyslog=yes
#ForwardToKMsg=no
#ForwardToConsole=no
#TTYPath=/dev/console
#MaxLevelStore=debug
#MaxLevelSyslog=debug
#MaxLevelKMsg=notice
#MaxLevelConsole=info

You have to remove the # in front of SystemMaxUse to make it work, # means that the line of code is ignored.

I-sty · 2014-02-21 19:25:06

Daw11 wrote:

You have to remove the # in front of SystemMaxUse to make it work, # means that the line of code is ignored.

I know, and I did.
Now:
http://imagerz.com/QBZFWEtvAwMCA1JNQgVQ
My boot time is 4 sec kernel + 1 sec grub + 30 sec systemd => 35~36 sec
Systemd says 18 sec (with userspace and with kernel).

drcouzelis · 2014-02-21 20:01:12

I-sty wrote:

My boot time is 4 sec kernel + 1 sec grub + 30 sec systemd => 35~36 sec
Systemd says 18 sec (with userspace and with kernel).

Does that mean we've won? What is your boot time goal?

I can't wait to get home and check my journal size...

I-sty · 2014-02-21 20:25:48

drcouzelis wrote:

I-sty wrote:
My boot time is 4 sec kernel + 1 sec grub + 30 sec systemd => 35~36 sec
Systemd says 18 sec (with userspace and with kernel).
Does that mean we've won? What is your boot time goal?
I can't wait to get home and check my journal size...

5sec kernel and 2sec userspace. My longest process is ~2 sec by the blame.
I am using i3 so my xorg loading is ~0.
In virtualbox I had 1.2 sec kernel and 0.8 sec userspace. :-d i3 loaded instant ;-)

Edit:
I use btrfs system with preload and readahead services

Last edited by I-sty (2014-02-21 20:31:04)

orschiro · 2014-02-22 06:22:44

@I-sty

So, solved by limiting the journal size?

I-sty · 2014-02-22 16:38:11

orschiro wrote:

@I-sty
So, solved by limiting the journal size?

Unfortunately not

The last power on, I have 6 sec kernel + 28 sec userspace but the real time is ~45 seconds.
I have a big "pause" in my userspace graph. The pause takes ~15 sec

I don't understand the systemd's logic.

http://imagerz.com/QBZFWUtvAwMCA1IYGQVQ

Last edited by I-sty (2014-02-22 16:43:47)

js · 2014-02-22 22:01:36

This sounds similar to a problem I had before. I found the only solutions were to either disable readahead through systemd or to have a nonpersistant journal.

I-sty · 2014-02-23 06:29:53

js wrote:

This sounds similar to a problem I had before. I found the only solutions were to either disable readahead through systemd or to have a nonpersistant journal.

Thanks js!,
I disabled the readahead service, and changed the journald storage from persistent to voilate

My "pause" in the plot file disappeared.
Now systemd says 4 sec kernel + 8 sec userspace, it would be very good, but unfortunately the real time is 30 sec.

Do the Xorg (I use i3) loads 18sec?

Last edited by I-sty (2014-02-23 16:49:19)

Spider.007 · 2014-02-23 11:58:25

Could you try disabling read-ahead without a nonpersistent journal? It might be interesting to know which of them causes this. Maybe you could also report this upstream and see if they have any debug-ideas. They might be able to fix this (for everyone to enjoy)

Thorsten Reinbold · 2014-02-23 16:28:04

I-sty wrote:

js wrote:
This sounds similar to a problem I had before. I found the only solutions were to either disable readahead through systemd or to have a nonpersistant journal.
Thanks js!,
I disabled the readahead service, and changed the journald storage from persistent to voilate
My "pause" in the plot file disappeared.
Now systemd says 4 sec kernel + 8 sec userspace, it would be very good, but unfortunately the real time is 30 sec.
The Xorg (I use i3) loads 18sec?

Beside the hints depending systemd: if you are using ext4 with a hdd, I suggest to have a look at e4rat. You can greatly decrease your boot time with this. Before I've got an ssd it worked very well for me.

Last edited by Thorsten Reinbold (2014-02-23 16:29:29)

I-sty · 2014-02-23 16:49:48

Thorsten Reinbold wrote:

Beside the hints depending systemd: if you are using ext4 with a hdd, I suggest to have a look at e4rat. You can greatly decrease your boot time with this. Before I've got an ssd it worked very well for me.

I use btrfs system

Arch Linux

#1 2014-02-20 15:22:59

How can I decrease my boot time? (60+ sec)

#2 2014-02-20 15:28:00

Re: How can I decrease my boot time? (60+ sec)

#3 2014-02-20 15:33:09

Re: How can I decrease my boot time? (60+ sec)

#4 2014-02-20 16:05:17

Re: How can I decrease my boot time? (60+ sec)

#5 2014-02-20 16:19:23

Re: How can I decrease my boot time? (60+ sec)

#6 2014-02-20 16:32:21

Re: How can I decrease my boot time? (60+ sec)

#7 2014-02-20 17:02:35

Re: How can I decrease my boot time? (60+ sec)

#8 2014-02-20 17:54:16

Re: How can I decrease my boot time? (60+ sec)

#9 2014-02-20 20:30:35

Re: How can I decrease my boot time? (60+ sec)

#10 2014-02-21 16:07:32

Re: How can I decrease my boot time? (60+ sec)

#11 2014-02-21 18:07:46

Re: How can I decrease my boot time? (60+ sec)

#12 2014-02-21 18:18:55

Re: How can I decrease my boot time? (60+ sec)

#13 2014-02-21 18:48:28

Re: How can I decrease my boot time? (60+ sec)

#14 2014-02-21 18:54:32

Re: How can I decrease my boot time? (60+ sec)

#15 2014-02-21 19:17:49

Re: How can I decrease my boot time? (60+ sec)

#16 2014-02-21 19:25:06

Re: How can I decrease my boot time? (60+ sec)

#17 2014-02-21 20:01:12

Re: How can I decrease my boot time? (60+ sec)

#18 2014-02-21 20:25:48

Re: How can I decrease my boot time? (60+ sec)

#19 2014-02-22 06:22:44

Re: How can I decrease my boot time? (60+ sec)

#20 2014-02-22 16:38:11

Re: How can I decrease my boot time? (60+ sec)

#21 2014-02-22 22:01:36

Re: How can I decrease my boot time? (60+ sec)

#22 2014-02-23 06:29:53

Re: How can I decrease my boot time? (60+ sec)

#23 2014-02-23 11:58:25

Re: How can I decrease my boot time? (60+ sec)

#24 2014-02-23 16:28:04

Re: How can I decrease my boot time? (60+ sec)

#25 2014-02-23 16:49:48

Re: How can I decrease my boot time? (60+ sec)

Board footer