docs/AUTOMATIC_BOOT_ASSESSMENT.md

   1 ---
   2 title: Automatic Boot Assessment
   3 ---
   4
   5 # Automatic Boot Assessment
   6
   7 systemd provides support for automatically reverting back to the previous
   8 version of the OS or kernel in case the system consistently fails to boot. This
   9 support is built into various of its components. When used together these
  10 components provide a complete solution on UEFI systems, built as add-on to the
  11 [Boot Loader
  12 Specification](https://systemd.io/BOOT_LOADER_SPECIFICATION). However, the
  13 different components may also be used independently, and in combination with
  14 other software, to implement similar schemes, for example with other boot
  15 loaders or for non-UEFI systems. Here's a brief overview of the complete set of
  16 components:
  17
  18 * The
  19   [`systemd-boot(7)`](https://www.freedesktop.org/software/systemd/man/systemd-boot.html)
  20   boot loader optionally maintains a per-boot-loader-entry counter that is
  21   decreased by one on each attempt to boot the entry, prioritizing entries that
  22   have non-zero counters over those which already reached a counter of zero
  23   when choosing the entry to boot.
  24
  25 * The
  26   [`systemd-bless-boot.service(8)`](https://www.freedesktop.org/software/systemd/man/systemd-bless-boot.service.html)
  27   service automatically marks a boot loader entry, for which boot counting as
  28   mentioned above is enabled, as "good" when a boot has been determined to be
  29   successful, thus turning off boot counting for it.
  30
  31 * The
  32   [`systemd-bless-boot-generator(8)`](https://www.freedesktop.org/software/systemd/man/systemd-bless-boot-generator.html)
  33   generator automatically pulls in `systemd-bless-boot.service` when use of
  34   `systemd-boot` with boot counting enabled is detected.
  35
  36 * The
  37   [`systemd-boot-check-no-failures.service(8)`](https://www.freedesktop.org/software/systemd/man/systemd-boot-check-no-failures.service.html)
  38   service is a simple health check tool that determines whether the boot
  39   completed successfully. When enabled it becomes an indirect dependency of
  40   `systemd-bless-boot.service` (by means of `boot-complete.target`, see
  41   below), ensuring that the boot will not be considered successful if there are
  42   any failed services.
  43
  44 * The `boot-complete.target` target unit (see
  45   [`systemd.special(7)`](https://www.freedesktop.org/software/systemd/man/systemd.special.html))
  46   serves as a generic extension point both for units that shall be considered
  47   necessary to consider a boot successful on one side (example:
  48   `systemd-boot-check-no-failures.service` as described above), and units that
  49   want to act only if the boot is successful on the other (example:
  50   `systemd-bless-boot.service` as described above).
  51
  52 * The
  53   [`kernel-install(8)`](https://www.freedesktop.org/software/systemd/man/kernel-install.html)
  54   script can optionally create boot loader entries that carry an initial boot
  55   counter (the initial counter is configurable in `/etc/kernel/tries`).
  56
  57 # Details
  58
  59 The boot counting data `systemd-boot` and `systemd-bless-boot.service`
  60 manage is stored in the name of the boot loader entries. If a boot loader entry
  61 file name contains `+` followed by one or two numbers (if two numbers, then
  62 those need to be separated by `-`) right before the `.conf` suffix, then boot
  63 counting is enabled for it. The first number is the "tries left" counter
  64 encoding how many attempts to boot this entry shall still be made. The second
  65 number is the "tries done" counter, encoding how many failed attempts to boot
  66 it have already been made. Each time a boot loader entry marked this way is
  67 booted the first counter is decreased by one, and the second one increased by
  68 one. (If the second counter is missing, then it is assumed to be equivalent to
  69 zero.) If the "tries left" counter is above zero the entry is still considered
  70 for booting (the entry's state is considered to be "indeterminate"), as soon as
  71 it reached zero the entry is not tried anymore (entry state "bad"). If the boot
  72 attempt completed successfully the entry's counters are removed from the name
  73 (entry state "good"), thus turning off boot counting for the future.
  74
  75 ## Walkthrough
  76
  77 Here's an example walkthrough of how this all fits together.
  78
  79 1. The user runs `echo 3 > /etc/kernel/tries` to enable boot counting.
  80
  81 2. A new kernel is installed. `kernel-install` is used to generate a new boot
  82    loader entry file for it. Let's say the version string for the new kernel is
  83    `4.14.11-300.fc27.x86_64`, a new boot loader entry
  84    `/boot/loader/entries/4.14.11-300.fc27.x86_64+3.conf` is hence created.
  85
  86 3. The system is booted for the first time after the new kernel is
  87    installed. The boot loader now sees the `+3` counter in the entry file
  88    name. It hence renames the file to `4.14.11-300.fc27.x86_64+2-1.conf`
  89    indicating that at this point one attempt has started and thus only one less
  90    is left. After the rename completed the entry is booted as usual.
  91
  92 4. Let's say this attempt to boot fails. On the following boot the boot loader
  93    will hence see the `+2-1` tag in the name, and hence rename the entry file to
  94    `4.14.11-300.fc27.x86_64+1-2.conf`, and boot it.
  95
  96 5. Let's say the boot fails again. On the subsequent boot the loader hence will
  97    see the `+1-2` tag, and rename the file to
  98    `4.14.11-300.fc27.x86_64+0-3.conf` and boot it.
  99
 100 6. If this boot also fails, on the next boot the boot loader will see the the
 101    tag `+0-3`, i.e. the counter reached zero. At this point the entry will be
 102    considered "bad", and ordered to the end of the list of entries. The next
 103    newest boot entry is now tried, i.e. the system automatically reverted back
 104    to an earlier version.
 105
 106 The above describes the walkthrough when the selected boot entry continuously
 107 fails. Let's have a look at an alternative ending to this walkthrough. In this
 108 scenario the first 4 steps are the same as above:
 109
 110 1. *as above*
 111
 112 2. *as above*
 113
 114 3. *as above*
 115
 116 4. *as above*
 117
 118 5. Let's say the second boot succeeds. The kernel initializes properly, systemd
 119    is started and invokes all generators.
 120
 121 6. One of the generators started is `systemd-bless-boot-generator` which
 122    detects that boot counting is used. It hence pulls
 123    `systemd-bless-boot.service` into the initial transaction.
 124
 125 7. `systemd-bless-boot.service` is ordered after and `Requires=` the generic
 126    `boot-complete.target` unit. This unit is hence also pulled into the initial
 127    transaction.
 128
 129 8. The `boot-complete.target` unit is ordered after and pulls in various units
 130    that are required to succeed for the boot process to be considered
 131    successful. One such unit is `systemd-boot-check-no-failures.service`.
 132
 133 9. `systemd-boot-check-no-failures.service` is run after all its own
 134    dependencies completed, and assesses that the boot completed
 135    successfully. It hence exits cleanly.
 136
 137 10. This allows `boot-complete.target` to be reached. This signifies to the
 138     system that this boot attempt shall be considered successful.
 139
 140 11. Which in turn permits `systemd-bless-boot.service` to run. It now
 141     determines which boot loader entry file was used to boot the system, and
 142     renames it dropping the counter tag. Thus
 143     `4.14.11-300.fc27.x86_64+1-2.conf` is renamed to
 144     `4.14.11-300.fc27.x86_64.conf`. From this moment boot counting is turned
 145     off.
 146
 147 12. On the following boot (and all subsequent boots after that) the entry is
 148     now seen with boot counting turned off, no further renaming takes place.
 149
 150 # How to adapt this scheme to other setups
 151
 152 Of the stack described above many components may be replaced or augmented. Here
 153 are a couple of recommendations.
 154
 155 1. To support alternative boot loaders in place of `systemd-boot` two scenarios
 156    are recommended:
 157
 158     a. Boot loaders already implementing the Boot Loader Specification can simply
 159        implement an equivalent file rename based logic, and thus integrate fully
 160        with the rest of the stack.
 161
 162     b. Boot loaders that want to implement boot counting and store the counters
 163        elsewhere can provide their own replacements for
 164        `systemd-bless-boot.service` and `systemd-bless-boot-generator`, but should
 165        continue to use `boot-complete.target` and thus support any services
 166        ordered before that.
 167
 168 2. To support additional components that shall succeed before the boot is
 169    considered successful, simply place them in units (if they aren't already)
 170    and order them before the generic `boot-complete.target` target unit,
 171    combined with `Requires=` dependencies from the target, so that the target
 172    cannot be reached when any of the units fail. You may add any number of
 173    units like this, and only if they all succeed the boot entry is marked as
 174    good. Note that the target unit shall pull in these boot checking units, not
 175    the other way around.
 176
 177 3. To support additional components that shall only run on boot success, simply
 178    wrap them in a unit and order them after `boot-complete.target`, pulling it
 179    in.
 180
 181 # FAQ
 182
 183 1. *Why do you use file renames to store the counter? Why not a regular file?*
 184    — Mainly two reasons: it's relatively likely that renames can be implemented
 185    atomically even in simpler file systems, while writing to file contents has
 186    a much bigger chance to be result in incomplete or corrupt data, as renaming
 187    generally avoids allocating or releasing data blocks. Moreover it has the
 188    benefit that the boot count metadata is directly attached to the boot loader
 189    entry file, and thus the lifecycle of the metadata and the entry itself are
 190    bound together. This means no additional clean-up needs to take place to
 191    drop the boot loader counting information for an entry when it is removed.
 192
 193 2. *Why not use EFI variables for storing the boot counter?* — The memory chips
 194    used to back the persistent EFI variables are generally not of the highest
 195    quality, hence shouldn't be written to more than necessary. This means we
 196    can't really use it for changes made regularly during boot, but can use it
 197    only for seldom made configuration changes.
 198
 199 3. *I have a service which — when it fails — should immediately cause a
 200    reboot. How does that fit in with the above?* — Well, that's orthogonal to
 201    the above, please use `FailureAction=` in the unit file for this.
 202
 203 4. *Under some condition I want to mark the current boot loader entry as bad
 204    right-away, so that it never is tried again, how do I do that?* — You may
 205    invoke `/usr/lib/systemd/systemd-bless-boot bad` at any time to mark the
 206    current boot loader entry as "bad" right-away so that it isn't tried again
 207    on later boots.