README.md

   1 # lxcfs
   2
   3 ## Introduction
   4 LXCFS is a small FUSE filesystem written with the intention of making Linux
   5 containers feel more like a virtual machine. It started as a side-project of
   6 `LXC` but is useable by any runtime.
   7
   8 LXCFS will take care that the information provided by crucial files in `procfs`
   9 such as:
  10
  11 ```
  12 /proc/cpuinfo
  13 /proc/diskstats
  14 /proc/meminfo
  15 /proc/stat
  16 /proc/swaps
  17 /proc/uptime
  18 /proc/slabinfo
  19 /sys/devices/system/cpu/online
  20 ```
  21
  22 are container aware such that the values displayed (e.g. in `/proc/uptime`)
  23 really reflect how long the container is running and not how long the host is
  24 running.
  25
  26 Prior to the implementation of cgroup namespaces by Serge Hallyn `LXCFS` also
  27 provided a container aware `cgroupfs` tree. It took care that the container
  28 only had access to cgroups underneath it's own cgroups and thus provided
  29 additional safety. For systems without support for cgroup namespaces `LXCFS`
  30 will still provide this feature but it is mostly considered deprecated.
  31
  32 ## Upgrading `LXCFS` without restart
  33
  34 `LXCFS` is split into a shared library (a libtool module, to be precise)
  35 `liblxcfs` and a simple binary `lxcfs`. When upgrading to a newer version of
  36 `LXCFS` the `lxcfs` binary will not be restarted. Instead it will detect that
  37 a new version of the shared library is available and will reload it using
  38 `dlclose(3)` and `dlopen(3)`. This design was chosen so that the fuse main loop
  39 that `LXCFS` uses will not need to be restarted. If it were then all containers
  40 using `LXCFS` would need to be restarted since they would otherwise be left
  41 with broken fuse mounts.
  42
  43 To force a reload of the shared library at the next possible instance simply
  44 send `SIGUSR1` to the pid of the running `LXCFS` process. This can be as simple
  45 as doing:
  46
  47     rm /usr/lib64/lxcfs/liblxcfs.so # MUST to delete the old library file first
  48     cp liblxcfs.so /usr/lib64/lxcfs/liblxcfs.so # to place new library file
  49     kill -s USR1 $(pidof lxcfs) # reload
  50
  51 ### musl
  52
  53 To achieve smooth upgrades through shared library reloads `LXCFS` also relies
  54 on the fact that when `dlclose(3)` drops the last reference to the shared
  55 library destructors are run and when `dlopen(3)` is called constructors are
  56 run. While this is true for `glibc` it is not true for `musl` (See the section
  57 [Unloading libraries](https://wiki.musl-libc.org/functional-differences-from-glibc.html).).
  58 So users of `LXCFS` on `musl` are advised to restart `LXCFS` completely and all
  59 containers making use of it.
  60
  61 ## Building
  62
  63 In order to build LXCFS install fuse and the fuse development headers according
  64 to your distro. LXCFS prefers `fuse3` but does work with new enough `fuse2`
  65 versions:
  66
  67     git clone git://github.com/lxc/lxcfs
  68     cd lxcfs
  69     meson setup -Dinit-script=systemd --prefix=/usr build/
  70     meson compile -C build/
  71     sudo meson install -C build/
  72
  73 To build with sanitizers you have to specify `-Db_sanitize=...` option to `meson setup`.
  74 For example, to enable ASAN and UBSAN:
  75
  76     meson setup -Dinit-script=systemd --prefix=/usr build/ -Db_sanitize=address,undefined
  77     meson compile -C build/
  78
  79 ## Usage
  80 The recommended command to run lxcfs is:
  81
  82     sudo mkdir -p /var/lib/lxcfs
  83     sudo lxcfs /var/lib/lxcfs
  84
  85 A container runtime wishing to use `LXCFS` should then bind mount the
  86 approriate files into the correct places on container startup.
  87
  88 ### LXC
  89 In order to use lxcfs with systemd-based containers, you can either use
  90 LXC 1.1 in which case it should work automatically, or otherwise, copy
  91 the `lxc.mount.hook` and `lxc.reboot.hook` files (once built) from this tree to
  92 `/usr/share/lxcfs`, make sure it is executable, then add the
  93 following lines to your container configuration:
  94 ```
  95 lxc.mount.auto = cgroup:mixed
  96 lxc.autodev = 1
  97 lxc.kmsg = 0
  98 lxc.include = /usr/share/lxc/config/common.conf.d/00-lxcfs.conf
  99 ```
 100
 101 ## Using with Docker
 102
 103 ```
 104 docker run -it -m 256m --memory-swap 256m \
 105       -v /var/lib/lxcfs/proc/cpuinfo:/proc/cpuinfo:rw \
 106       -v /var/lib/lxcfs/proc/diskstats:/proc/diskstats:rw \
 107       -v /var/lib/lxcfs/proc/meminfo:/proc/meminfo:rw \
 108       -v /var/lib/lxcfs/proc/stat:/proc/stat:rw \
 109       -v /var/lib/lxcfs/proc/swaps:/proc/swaps:rw \
 110       -v /var/lib/lxcfs/proc/uptime:/proc/uptime:rw \
 111       -v /var/lib/lxcfs/proc/slabinfo:/proc/slabinfo:rw \
 112       -v /var/lib/lxcfs/sys/devices/system/cpu:/sys/devices/system/cpu:rw \
 113       ubuntu:18.04 /bin/bash
 114  ```
 115
 116  In a system with swap enabled, the parameter "-u" can be used to set all values in "meminfo" that refer to the swap to 0.
 117
 118  sudo lxcfs -u /var/lib/lxcfs
 119
 120 ## Swap handling
 121 If you noticed LXCFS not showing any SWAP in your container despite
 122 having SWAP on your system, please read this section carefully and look
 123 for instructions on how to enable SWAP accounting for your distribution.
 124
 125 Swap cgroup handling on Linux is very confusing and there just isn't a
 126 perfect way for LXCFS to handle it.
 127
 128 Terminology used below:
 129  - RAM refers to `memory.usage_in_bytes` and `memory.limit_in_bytes`
 130  - RAM+SWAP refers to `memory.memsw.usage_in_bytes` and `memory.memsw.limit_in_bytes`
 131
 132 The main issues are:
 133  - SWAP accounting is often opt-in and, requiring a special kernel boot
 134    time option (`swapaccount=1`) and/or special kernel build options
 135    (`CONFIG_MEMCG_SWAP`).
 136
 137  - Both a RAM limit and a RAM+SWAP limit can be set. The delta however
 138    isn't the available SWAP space as the kernel is still free to SWAP as
 139    much of the RAM as it feels like. This makes it impossible to render
 140    a SWAP device size as using the delta between RAM and RAM+SWAP for that
 141    wouldn't account for the kernel swapping more pages, leading to swap
 142    usage exceeding swap total.
 143
 144  - It's impossible to disable SWAP in a given container. The closest
 145    that can be done is setting swappiness down to 0 which severly limits
 146    the risk of swapping pages but doesn't eliminate it.
 147
 148 As a result, LXCFS had to make some compromise which go as follow:
 149  - When SWAP accounting isn't enabled, no SWAP space is reported at all.
 150    This is simply because there is no way to know the SWAP consumption.
 151    The container may very much be using some SWAP though, there's just
 152    no way to know how much of it and showing a SWAP device would require
 153    some kind of SWAP usage to be reported. Showing the host value would be
 154    completely wrong, showing a 0 value would be equallty wrong.
 155
 156  - Because SWAP usage for a given container can exceed the delta between
 157    RAM and RAM+SWAP, the SWAP size is always reported to be the smaller of
 158    the RAM+SWAP limit or the host SWAP device itself. This ensures that at no
 159    point SWAP usage will be allowed to exceed the SWAP size.
 160
 161  - If the swappiness is set to 0 and there is no SWAP usage, no SWAP is reported.
 162    However if there is SWAP usage, then a SWAP device of the size of the
 163    usage (100% full) is reported. This provides adequate reporting of
 164    the memory consumption while preventing applications from assuming more
 165    SWAP is available.