zram-config

Overview

This is a complete zram-config utility for swap, directories, and logs to reduce SD, NAND and eMMC block wear. zram-config implements zram to prevent frequent writing to the disk and allow near ram speed access to working directories with varying compression ratios depending on the compression algorithm.

A ztab table in /etc/ztab is used to configure where any combination and number of zram drives are to be created. This project uses an OverlayFS mount with zram so that syncFromDisk on start is not needed. In theory this should allow for faster boots and larger directories as no complete directory copy is needed as it is the lower mount in the OverlayFS. Many thanks go to @kmxz for the overlayfs-tools used to make this possible.

The rationale for zram-config is that many distributions have zram-config packages that are actually broken, even by name, as often they are a zram-swap-config package in reality. But even then they do not check for other zram services or change the parameters of swap from HD based configurations to ram optimized ones. If all you are looking for is a zram-swap utility see StuartIanNaylor/zram-swap-config.

Both StuartIanNaylor/zram-swap-config and ecdye/zram-config are great examples for distributions to get their zram packages updated.

Also if the OverlayFS guys would actually make some official merge/snapshot tools and not just leave it as just enough for Docker that would be massively useful, and if anyone fancies shouting out that call please do.

A Brief Usage Guide

Table of Contents

1. Install
2. Update
3. Uninstall
4. Configure
   • Example configuration
5. Is it working?
6. Performance
7. Reference

Install

sudo apt-get install git
git clone --recurse-submodules https://github.com/ecdye/zram-config
cd zram-config
sudo ./install.bash

Note: The recommended way to stop the zram-config.service is to run

sudo zram-config "stop"

NOT

sudo systemctl stop zram-config.service

because of issues with the way systemd works with zram logging.

The service will stop normally on reboot, there is no need to manually stop it.

Update

cd /path/to/zram-config/
sudo ./update.bash

Uninstall

sudo /usr/local/share/zram-config/uninstall.bash

Configure

All configuration is done in the /etc/ztab file.

Use # to comment out any line, add new drives with the first column providing the drive type and then drive details separated by tab characters.

All algorithms in /proc/crypto are supported but only lzo-rle, lzo, lz4, and zstd have zramctl text strings; lzo-rle is the fastest with zstd having much better text compression.

mem_limit is the compressed memory limit and will set a hard memory limit for the system admin.

disk_size is the virtual uncompressed size approx. 220-450% of memory allocated depending on the algorithm and input file. Don't make it much higher than the compression algorithm is capable of as it will waste memory because there is a ~0.1% memory overhead when empty.

swap_priority will set zram over alternative swap devices.

page-cluster 0 means tuning to singular pages rather than the default 3 which caches 8 for HDD tuning, which can lower latency.

swappiness 80 because the improved performance of zram allows more usage without any adverse affects from the default of 60. It can be raised up to 100 but that will increase process queue on intense loads such as boot time.

target_dir is the directory you wish to hold in zram, and the original will be moved to a bind mount bind_dir and is synchronized on start, stop, and write commands.

bind_dir is the directory where the original directory will be mounted for sync purposes. Usually in /opt or /var, name optional.

oldlog_dir will enable log-rotation to an off device directory while retaining only live logs in zram. Usually in /opt or /var, name optional.

If you need multiple zram swaps or zram directories, just create another entry in /etc/ztab. To do this safely, first stop zram using sudo zram-config "stop", then edit /etc/ztab. Once finished, restart zram using sudo systemctl start zram-config.service.

Example configuration

# swap	alg		mem_limit	disk_size	swap_priority	page-cluster	swappiness
swap	lzo-rle		250M		750M		75		0		80

# dir	alg		mem_limit	disk_size	target_dir		bind_dir
#dir	lzo-rle		50M		150M		/home/pi		/pi.bind

# log	alg		mem_limit	disk_size	target_dir		bind_dir		oldlog_dir
log	lzo-rle		50M		150M		/var/log		/log.bind		/opt/zram/oldlog

Is it working?

Run zramctl in your preferred shell and if you see and output similar to below, yes it is working.

[email protected]:~ $ zramctl
NAME       ALGORITHM DISKSIZE  DATA COMPR TOTAL STREAMS MOUNTPOINT
/dev/zram0 lz4           1.2G    4K   76B    4K       4 [SWAP]
/dev/zram1 lz4           150M 16.3M 25.1K  208K       4 /opt/zram/zram1
/dev/zram2 lz4            60M  7.5M  1.2M  1.7M       4 /opt/zram/zram2

To view more information on zram usage take a look at the following commands and their corresponding output.

[email protected]:~ $ df
Filesystem     1K-blocks    Used Available Use% Mounted on
/dev/root       14803620 2558152  11611220  19% /
devtmpfs          470116       0    470116   0% /dev
tmpfs             474724  223868    250856  48% /dev/shm
tmpfs             474724   12284    462440   3% /run
tmpfs               5120       4      5116   1% /run/lock
tmpfs             474724       0    474724   0% /sys/fs/cgroup
/dev/mmcblk0p1     44220   22390     21831  51% /boot
/dev/zram1        132384     280    121352   1% /opt/zram/zram1
overlay1          132384     280    121352   1% /home/pi/MagicMirror
/dev/zram2         55408    3460     47648   7% /opt/zram/zram2
overlay2           55408    3460     47648   7% /var/log
tmpfs              94944       0     94944   0% /run/user/1000

[email protected]:~ $ free -h
              total        used        free      shared  buff/cache   available
Mem:           927M        206M        184M        233M        535M        434M
Swap:          1.3G          0B        1.3G

[email protected]:~ $ swapon
NAME       TYPE      SIZE USED PRIO
/dev/zram0 partition 1.2G   0B   75
/var/swap  file      100M   0B   -2

[email protected]:/opt/zram $ ls
log.bind  magicmirror.bind  oldlog  zram1  zram2

[email protected]:/opt/zram $ top
top - 23:18:21 up  1:28,  2 users,  load average: 0.31, 0.29, 0.29
Tasks: 114 total,   1 running,  68 sleeping,   0 stopped,   0 zombie
%Cpu(s):  1.9 us,  0.1 sy,  0.0 ni, 98.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem :   949448 total,   153464 free,   223452 used,   572532 buff/cache
KiB Swap:  1331192 total,  1331192 free,        0 used.   412052 avail Mem

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
 1215 pi        20   0  600844 325968 287276 S   5.3 34.3   8:09.51 chromium-browse
 2536 pi        20   0    8104   3204   2728 R   1.6  0.3   0:00.11 top
  970 pi        20   0  775108 156128 112876 S   1.0 16.4  11:17.06 chromium-browse
 1611 pi        20   0   11656   3772   3056 S   0.3  0.4   0:00.30 sshd
    1 root      20   0   27072   5964   4824 S   0.0  0.6   0:02.51 systemd
    2 root      20   0       0      0      0 S   0.0  0.0   0:00.00 kthreadd
    4 root       0 -20       0      0      0 I   0.0  0.0   0:00.00 kworker/0:0H
    6 root       0 -20       0      0      0 I   0.0  0.0   0:00.00 mm_percpu_wq
    7 root      20   0       0      0      0 S   0.0  0.0   0:00.24 ksoftirqd/0
    8 root      20   0       0      0      0 I   0.0  0.0   0:00.87 rcu_sched
    9 root      20   0       0      0      0 I   0.0  0.0   0:00.00 rcu_bh

Performance

LZO-RLE offers the best performance and is probably the best choice, and from kernel 5.1 and onward it is the default. You might have text based low impact directories such as /var/log or /var/cache where a highly effective text compressor such as zstd is optimal, with effective compression that can be up to 200% of what LZO may achieve especially with text. With /tmp and /run, zram is unnecessary because they are RAM mounted as tmpfs and, if memory gets short, then the zram swap will provide extra. It is only under intense loads that the slight overhead of zram compression becomes noticeable.

This chart from facebook/zstd provides a good benchmark for the performance of the different compressors.

With swap, zram changes what is normally a static assumption that a HD is providing the swap using swapiness and page-cache where default swapiness is 60 and page-cache is 3. Depending on the average load zram will benefit from a setting of 80-100 for swapiness and changing page-cache to 0 so that singular pages are written which will greatly reduce latency. It is a shame swapiness is not dynamically based on load as for many systems there is often a huge difference in boot startup to settled load. In some cases you may find you are reducing swapiness purely because of boot load.

Reference

https://www.kernel.org/doc/Documentation/blockdev/zram.txt

https://www.kernel.org/doc/Documentation/filesystems/overlayfs.txt