Announcement

**GreyGeek** · Oct 22, 2018, 07:51 PM

Personally, I'd go with the whole device. Btrfs includes native optimizations for SSD media. To enable these features, mount the Btrfs filesystem with the -o ssd mount option. These optimizations include enhanced SSD write performance by avoiding optimization such as seek optimizations which do not apply to solid-state media. And, have a cron job to do an fstrim once a week and another to do a balance once a week.

**Thomas00** · Oct 23, 2018, 01:22 AM

Swap may require additional attention. If you want to use all of sdc for btrfs you have to create or reuse a swap partition on a different drive. Or not use swap at all...

https://btrfs.wiki.kernel.org/index.php/FAQ
Does btrfs support swap files?
Currently no...

Not sure about the cron job for trim, on my Ubuntu 18.04 server and my Kubuntu 18.10 laptop the installer set up trim automatically.

Code:

thomas@hermes:~$ sudo systemctl status fstrim.timer
[sudo] password for thomas: 
● fstrim.timer - Discard unused blocks once a week
Loaded: loaded (/lib/systemd/system/fstrim.timer; enabled; vendor preset: enabled)
Active: active (waiting) since Sun 2018-10-21 21:55:52 CEST; 1 day 11h ago
Trigger: Mon 2018-10-29 00:00:00 CET; 5 days left
  Docs: man:fstrim

Oct 21 21:55:52 hermes systemd[1]: Started Discard unused blocks once a week.

**vinnywright** · Oct 23, 2018, 07:19 PM

OK so just did it ,,,,

Code:

vinny@vinny-Bonobo-Extreme:~$ sudo mkfs.btrfs /dev/sdc
[sudo] password for vinny: 
btrfs-progs v4.15.1
See http://btrfs.wiki.kernel.org for more information.

Detected a SSD, turning off metadata duplication. Mkfs with -m dup if you want to force metadata duplication.
Performing full device TRIM /dev/sdc (232.89GiB) ...
Label:              (null)
UUID:               fe385e89-57a5-4632-9823-043a70b67c65
Node size:          16384
Sector size:        4096
Filesystem size:    232.89GiB
Block group profiles:
Data:             single            8.00MiB
Metadata:         single            8.00MiB
System:           single            4.00MiB
SSD detected:       yes
Incompat features:  extref, skinny-metadata
Number of devices:  1
Devices:
ID        SIZE  PATH
 1   232.89GiB  /dev/sdc

used Kubuntu-18.04 as the first install .
installed with no grub as the Neon-/dev/stable in the 500GB HD is controlling the grub you see at startup .
the ssd and Kubuntu gets booted with this custom entry in Neon-/dev/stable's /etc/grub.d/40_custom ,,,

Code:

menuentry 'Kubuntu-18.04' {
insmod btrfs
set root='(hd2)'
linux /@/vmlinuz root=UUID=fe385e89-57a5-4632-9823-043a70b67c65 rootflags=subvol=@ ro quiet splash
initrd /@/initrd.img
}

I use the "vmlinuz" & "initrd.img" links in / as their always updated to the new Kernel and image when you get one , so no need to do any update-grub'ing in the install that controls grub to boot your new stuff.

the thing boots in about 3sec from grub to login and login to a usable desktop is kinda just a fade in after pressing enter ,,,,,,nice
no bouncing firefox ball either at first launch it just opens rite up .

on first boot their was a 390+ package upgrade wating for me , took 3-4min never seen the unpacking going by so quick .

trim seems to be setup already

Code:

vinny@vinny-bonox8:~$ sudo systemctl status fstrim.timer
[sudo] password for vinny: 
● fstrim.timer - Discard unused blocks once a week
Loaded: loaded (/lib/systemd/system/fstrim.timer; enabled; vendor preset: enabled)
Active: active (waiting) since Tue 2018-10-23 20:19:55 EDT; 31min ago
Trigger: Mon 2018-10-29 00:00:00 EDT; 5 days left
  Docs: man:fstrim

Oct 23 20:19:55 vinny-bonox8 systemd[1]: Stopping Discard unused blocks once a week.
Oct 23 20:19:55 vinny-bonox8 systemd[1]: Started Discard unused blocks once a week.

any one else that has some ideas for SSD optimizations pleas chime in

all for now ,,,,,gotta get to bead for work , 4:00AM comes early ,,,,,LOL

VINNY

EDIT: OK I just actually timed the boot ,,11sec and 4sec to desktop

**jlittle** · Oct 23, 2018, 11:30 PM

set root='(hd2)'

That prompts some observations. Here I assume that grub menuentry works.

The grub set root statement I'm used to seeing has the partition, say set root='hd2,gpt2', but I suppose if there's no partition table there's no partition. However, in the grub.cfg files generated by grub-mkconfig the variable is set without the parentheses. grub's syntax is weird and forgiving here, on my system all these are accepted and work:
Code:
```
set root=hd2,gpt2
set root='hd2,gpt2'
set root=(hd2,gpt2)
set root='(hd2,gpt2)'
set root="hd2,gpt2"
set root="(hd2,gpt2)"
```
I have to try hard to get something that doesn't:set root=((hd2,gpt2)). To avoid confusion I think it's better to use the form that grub uses, which in your case would be set root='hd2'.
Relying on a fixed device name like that may cause problems down the line. The way grub enumerates its devices can change from boot to boot unpredictably, especially when removable devices are plugged in. Grub normally has the set root statement like that as a fallback, and $root is set by a search statement that depends on something that won't change. In your case that could be
Code:
```
search --no-floppy --fs-uuid --set=root fe385e89-57a5-4632-9823-043a70b67c65
```
I find those UUIDs horrible to read, and prefer to label my partitions in a way I understand, and use the labels. I boot bionic with
Code:
```
set root='hd2,gpt2'
search --no-floppy --set=root --label "main"
linux /@_bionic/vmlinuz root=PARTLABEL=main ro rootflags=subvol=@_bionic
initrd /@_bionic/initrd.img
```
I'm not sure if whole devices can have a label, though. The linux line can be root=LABEL=main; I boot cosmic with that, it doesn't seem to make a difference; but that difference would seem to apply to the partitionless situation.
Committing to booting from another disk lessens resilience, I think; if that disk fails, no boot.
What are the advantages to the partitionless btrfs?
It's not crucial, however. Just have something else with a bootloader (such as grub or syslinux) on it, say a removable drive or stick, and boot from that. But, know where it is, and the commands to use.

...boots in about 3sec from grub to login...

Isn't it fantastic? It made me begrudge the 12 s or so for the hardware boot and POST, so I use suspend a lot now.

**GreyGeek** · Oct 24, 2018, 06:02 AM

I second what jlittle says about avoiding the use of /dev/sdX identifiers in scripts that mount drives. Removebor plug in another drive and the sdX designations WILL. I’ve experienced this while running a Btrfs RAID1 setup. It was built using sda & sdb but when I plugged in my 3rd HD instead of becoming sdc it became sdb!!! My RAID1 then became sda & sdc. Worse, Btrfs started throwing errors saying that sdb & sdc had the same labels. That error didn’t hurt or change RAID or Btrfs performance but I had to keep an eye on the size of the log and trim it occasionally.

I’m not adverse to using UUID’s because the blkid command displays them easily and copy & paste is trivial. Also, while doing Btrfs maintenance I usually use the CLI history to recall mount, snapshot and send commands, which is why I haven’t bothered to write a script to do my snapshotting.

**vinnywright** · Oct 24, 2018, 06:31 PM

Originally posted by jlittle View Post

Committing to booting from another disk lessens resilience, I think; if that disk fails, no boot.
What are the advantages to the partitionless btrfs?
It's not crucial, however. Just have something else with a bootloader (such as grub or syslinux) on it, say a removable drive or stick, and boot from that. But, know where it is, and the commands to use.

Isn't it fantastic? It made me begrudge the 12 s or so for the hardware boot and POST, so I use suspend a lot now.

how is that ?

grub installs to 1 place usually !

IF you have 1 drive their is 1 grub install in the MBR that loads up when you start the computer , yes you can have several installs in different partitions and they can have their own grub installs in the PBR BUT the grub install in the MBR must chainload/point to them , so still their is only 1 grub that starts the process.

now if you have more than one drive you can have a grub install in each one of their MBR's BUT to get to it you will half to tell the BIOS/UEFI to boot that HD .

@hear I do it like this ,,,,,,,

Code:

vinny@vinny-bonox8:~$ sudo parted -l
[sudo] password for vinny: 
Model: ATA HGST HTS725050A7 (scsi)
Disk /dev/sda: 500GB
Sector size (logical/physical): 512B/4096B
Partition Table: msdos
Disk Flags: 

Number  Start   End    Size    Type      File system     Flags
1      8225kB  323GB  323GB   primary   btrfs           boot
3      323GB   379GB  56.3GB  primary   ext4
4      379GB   496GB  117GB   extended
6      379GB   436GB  57.0GB  logical   ext4
5      436GB   496GB  59.8GB  logical   ext4
2      496GB   500GB  4295MB  primary   linux-swap(v1)

Model: ATA HGST HTS721010A9 (scsi)
Disk /dev/sdb: 1000GB
Sector size (logical/physical): 512B/4096B
Partition Table: gpt
Disk Flags: 

Number  Start   End     Size    File system  Name     Flags
1      1049kB  1000GB  1000GB  btrfs        primary

Model: ATA Samsung SSD 860 (scsi)
Disk /dev/sdc: 250GB
Sector size (logical/physical): 512B/512B
Partition Table: loop
Disk Flags: 

Number  Start  End    Size   File system  Flags
1      0.00B  250GB  250GB  btrfs

/dev/sda has the main grub that I see when I boot the box .
/dev/sda1 has 2 installs neon-lts and Kubuntu 18.04 the Kubuntu has a grub install in the PBR of sda1 neon has no grub install , IF I chose at the grub screen the Kubuntu in sda1 I get it's grub screen the neon-lts gets booted like I'm booting the Kubuntu in the SSD (/dev/sdc) .

grubs os-prober will not find installs in BTRFS so booting to one requires a custom menu entry unless it's grub is controlling things and even then it will not find OS's in other BTRFS drives/partitions/subvolumes

currently their are 6 installs in this box 5 on sda and 1(so far) in sdc , none of the 3 in BTRFS get found when Neon-/dev/stable (the OS that controls the main grub) gets "update-grub" ran , so they all have a custom menu entry .

it boils down like this

sda-has grub in the MBR contoled by Neon-/dev/stable (sda6)
sda1-BTRFS-Kubuntu-18.04 and Neon-LTS , Kubuntu has grub in the PBR of sda1
sda2-swap partition
sda3-EXT4-Kubuntu-14.04 , has grub in PBR of sda3
sda4-extended
sda5-EXT4-Debian8-KDE , grub in PBR of sda5
sda6-EXT4-Neon-/dev/stable , grub in MBR of sda

sdc-Kubuntu-18.04 hole disk BTRFS so no MBR or PBR

sdb-BTRFS-1TB storage drive that holds all my data that's shared with all OS's like this .

Code:

drwxr-xr-x 1 vinny vinny 70 Oct 23 20:07 Desktop
lrwxrwxrwx 1 vinny vinny 30 Oct 23 20:32 Documents -> /mnt/storage/kubuntu/Documents
lrwxrwxrwx 1 vinny vinny 30 Oct 23 20:32 Downloads -> /mnt/storage/kubuntu/Downloads
lrwxrwxrwx 1 vinny vinny 29 Oct 23 20:32 dwhelper -> /mnt/storage/kubuntu/dwhelper
lrwxrwxrwx 1 vinny vinny 26 Oct 23 20:33 Music -> /mnt/storage/kubuntu/Music
lrwxrwxrwx 1 vinny vinny 29 Oct 23 20:34 Pictures -> /mnt/storage/kubuntu/Pictures
drwxr-xr-x 1 vinny vinny  0 Oct 23 20:07 Public
drwxr-xr-x 1 vinny vinny  0 Oct 23 20:07 Templates
lrwxrwxrwx 1 vinny vinny 27 Oct 23 20:34 Videos -> /mnt/storage/kubuntu/Videos

every install has my home directory's linked to this 1 place

so what ever OS i'm in I have my stuff.

VINNY

**jlittle** · Oct 26, 2018, 05:15 PM

Originally posted by vinnywright View Post

how is that ?

grub installs to 1 place usually !

IF you have 1 drive their is 1 grub install in the MBR that loads up when you start the computer , yes you can have several installs in different partitions and they can have their own grub installs in the PBR BUT the grub install in the MBR must chainload/point to them , so still their is only 1 grub that starts the process.

I'm sorry, that condensed a fog of thickness in me. You're right for MBR, the grub installed in the MBR gap (or the MBR partition on a hybrid GPT) has to point somewhere. I was an early adopter of GPT and think a bit differently, especially since I divined how grub 2 works with EFI. But I'm suggesting that random failures of devices or OS's or grub are possible, and being able to boot from the firmware into the SSD might one day be useful.

grubs os-prober will not find installs in BTRFS so booting to one requires a custom menu entry unless it's grub is controlling things

I'm now right into grub custom menu entries, after my fun booting from isos directly. (Who needs to install an OS? Kubuntu runs really well in "Try Kubuntu" mode, other than pesky things like saving data and settings.) I've suggested that booting into Kubuntu on your sdb from grub on a USB stick would work well. I intend to test out some of the multi-boot projects; they promise to be able to boot any OS on any x86, 32 or 64 bit, out of the box.

**jlittle** · Oct 26, 2018, 05:31 PM

the thing boots in about 3sec from grub to login...
OK I just actually timed the boot ,,11sec and 4sec to desktop

Is that 11 s from power on to grub, and 4 s from grub menu to login?

Note that I've boasted about my 3 s boot in the past. Since I installed more of mysql and also postgresql it's become 4 s. Now that's a bit misleading in that systemd is running lots in parallel behind the scenes, and some things don't complete till a few seconds after I've reached the desktop, but I haven't notice any delays starting anything so effectively there's 4 s to login, 1 s to desktop.

**dibl** · Oct 27, 2018, 01:29 AM

Good use of $60, Vinny!

On this 4-year old laptop:

Code:

root@dibl-HP350:/# inxi -Fz
System:    Host: dibl-HP350 Kernel: 4.19.0-towo.1-siduction-amd64 x86_64 bits: 64 
         Desktop: KDE Plasma 5.13.5 
         Distro: siduction 18.3.0 Patience - kde - (201805132121) 
Machine:   Type: Laptop System: Hewlett-Packard product: HP 350 G1 
         v: 0992100000400010000634101 serial: <filter> 
         Mobo: Hewlett-Packard model: 21B7 v: KBC Version 46.0B serial: <filter> 
         UEFI: Insyde v: F.12 date: 06/12/2014 
Battery:   ID-1: BAT0 charge: 26.8 Wh condition: 31.4/31.4 Wh (100%) 
CPU:       Topology: Dual Core model: Intel Core i5-4210U bits: 64 type: MT MCP 
         L2 cache: 3072 KiB 
         Speed: 998 MHz min/max: 800/2700 MHz Core speeds (MHz): 1: 998 2: 999 3: 998 
         4: 998 
Graphics:  Device-1: Intel Haswell-ULT Integrated Graphics driver: i915 v: kernel 
         Display: x11 server: X.Org 1.20.3 driver: modesetting unloaded: fbdev,vesa 
         resolution: 1366x768~60Hz 
         OpenGL: renderer: Mesa DRI Intel Haswell Mobile v: 4.5 Mesa 18.1.9 
Audio:     Device-1: Intel Haswell-ULT HD Audio driver: snd_hda_intel 
         Device-2: Intel 8 Series HD Audio driver: snd_hda_intel 
         Sound Server: ALSA v: k4.19.0-towo.1-siduction-amd64 
Network:   Device-1: Realtek RTL8111/8168/8411 PCI Express Gigabit Ethernet driver: r8169 
         IF: eno1 state: down mac: <filter> 
         Device-2: Ralink RT3290 Wireless 802.11n 1T/1R PCIe driver: rt2800pci 
         IF: wlo1 state: up mac: <filter> 
Drives:    Local Storage: total: 465.76 GiB used: 8.35 GiB (1.8%) 
         ID-1: /dev/sda vendor: Samsung model: SSD 840 EVO 500GB size: 465.76 GiB 
Partition: ID-1: / size: 18.69 GiB used: 8.28 GiB (44.3%) fs: ext4 dev: /dev/sda2 
         ID-2: swap-1 size: 2.00 GiB used: 0 KiB (0.0%) fs: swap dev: /dev/sda3 
Sensors:   System Temperatures: cpu: 34.0 C mobo: N/A 
         Fan Speeds (RPM): N/A 
Info:      Processes: 183 Uptime: 34m Memory: 3.77 GiB used: 1.66 GiB (43.9%) Shell: bash 
         inxi: 3.0.27

I installed a Samsung 840 EVO right after I bought the laptop. It is partitioned like this:

Code:

Disk /dev/sda: 465.8 GiB, 500107862016 bytes, 976773168 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: gpt
Disk identifier: 1530E434-5805-49C0-9EC9-6A78D9E0EA3C

Device        Start       End   Sectors   Size Type
/dev/sda1      4097    618497    614401   300M EFI System
/dev/sda2    618498  42563583  41945086    20G Linux filesystem
/dev/sda3  42563584  46757887   4194304     2G Linux swap
/dev/sda4  46757888 976773119 930015232 443.5G Linux filesystem

In /etc/fstab I slow down the ext4 journaling to spare the SSD some wear, and also mount /tmp on a tmpfs:

Code:

# <file system>             <mount point>  <type>  <options>  <dump>  <pass>
UUID=8612-8993                            /boot/efi      vfat    defaults,noatime 0 2
UUID=325fc450-294d-4c11-a2f6-5e743fe8951e /              ext4    defaults,noatime,commit=120 0 1
UUID=a21b748d-1b48-4120-a8ac-f70a9f42a7d4 swap           swap    defaults,noatime 0 2
UUID=c8da0442-3b92-4787-abba-d30bf86838d1 /mnt/DATA      ext4    defaults,noatime,commit=120 0 2
tmpfs                                     /tmp           tmpfs   defaults,noatime,mode=1777 0 0

Booting to the Plasma desktop is pretty brisk:

Code:

root@dibl-HP350:/# systemd-analyze blame
        1.157s dev-disk-by\x2duuid-a21b748d\x2d1b48\x2d4120\x2da8ac\x2df70a9f42a7d4.swap
         746ms dev-sda2.device
         511ms systemd-journal-flush.service
         418ms systemd-logind.service
         324ms udisks2.service
         302ms minidlna.service
         290ms accounts-daemon.service
         272ms ModemManager.service
         266ms loadcpufreq.service
         201ms upower.service
         199ms NetworkManager.service
.
.
.

Another trick to (a) reduce cache-writing by the browser on the SSD, and also (b) let the browser cache go "poof" at shutdown, is a little configuration tweak on my SRware Iron browser, in "Command" properties:

Code:

/usr/share/iron/chrome-wrapper --disk-cache-dir=/run/user/1000/chromium-cache

I hope there's something useful in here somewhere.

**vinnywright** · Oct 27, 2018, 11:10 AM

Originally posted by dibl View Post

Good use of $60, Vinny!

In /etc/fstab I slow down the ext4 journaling to spare the SSD some wear, and also mount /tmp on a tmpfs:

Code:

# <file system>             <mount point>  <type>  <options>  <dump>  <pass>
UUID=8612-8993                            /boot/efi      vfat    defaults,noatime 0 2
UUID=325fc450-294d-4c11-a2f6-5e743fe8951e /              ext4    defaults,noatime,commit=120 0 1
UUID=a21b748d-1b48-4120-a8ac-f70a9f42a7d4 swap           swap    defaults,noatime 0 2
UUID=c8da0442-3b92-4787-abba-d30bf86838d1 /mnt/DATA      ext4    defaults,noatime,commit=120 0 2
tmpfs                                     /tmp           tmpfs   defaults,noatime,mode=1777 0 0

Booting to the Plasma desktop is pretty brisk:

Another trick to (a) reduce cache-writing by the browser on the SSD, and also (b) let the browser cache go "poof" at shutdown, is a little configuration tweak on my SRware Iron browser, in "Command" properties:

Code:

/usr/share/iron/chrome-wrapper --disk-cache-dir=/run/user/1000/chromium-cache

I hope there's something useful in here somewhere.

thanks for the reminder of setting "tmpfs" and "noatime" in "fstab" ,,I had forgotten to do those

firefox can be made to use "memory" for the cache , just did that as well .

VINNY

**vinnywright** · Oct 27, 2018, 07:00 PM

did a few more tweaks turning off some systemd.services and got the total time to desktop down to 9.5s

Code:

vinny@vinny-bonox8:~$ systemd-analyze 
Startup finished in 5.967s (kernel) + 3.477s (userspace) = 9.445s
graphical.target reached after 3.473s in userspace

VINNY

**jlittle** · Oct 27, 2018, 07:10 PM

Originally posted by vinnywright View Post

graphical.target reached after 3.473s in userspace

That's when you got the log in prompt, which I've argued above is the most relevant.

**vinnywright** · Oct 27, 2018, 07:32 PM

Code:

Startup finished in 5.967s (kernel)

is the time from grub to login

Code:

graphical.target reached after 3.473s in userspace

is time from login screen to desktop , as far as I can tell , sorry if that's the same thing you are saying I get confused easy , and could be wrong .

VINNY

**dibl** · Oct 28, 2018, 01:30 PM

Originally posted by jlittle View Post

That's when you got the log in prompt, which I've argued above is the most relevant.

Makes good sense. Here's the figure for the 840 EVO that I described above:

Code:

root@dibl-HP350:/home/don# systemd-analyze
Startup finished in 4.053s (kernel) + 2.880s (userspace) = 6.934s
graphical.target reached after 2.865s in userspace

Announcement

new SSD for OS's , guidance please

new SSD for OS's , guidance please

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment