about summary refs log tree commit diff
path: root/README.md
blob: 600bd08c16767eba16d7adee291875accfbce44e (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
## snooze: run a command at a particular time

`snooze` is a new tool for waiting until a particular time and then
running a command.  Together with a service supervision system such as
runit, this can be used to replace cron(8).

`snooze` has been tested on Linux 4.2.
It will likely work on other Unix-like systems with C99.

## Benefits

Over cron:
- mnemonic syntax
- no overlapping job runs possible
- filtering by ISO week and day of year
- due to supervision, no centralized daemon required
- due to supervision, can easily disable jobs or force their
  execution instantly
- due to supervision, have custom logs
- due to no centralized daemon, no fuzzing with multiple users/permissions
- very robust with respect to external time changes
- can use a file timestamp to ensure minimum waiting time between two
  runs, even across reboots
- randomized delays (some cron have that)
- variable slack (no need for anacron)
- ad-hoc usage possible, just run the program from command line

Over runwhen:
- less confusing usage (I hope)
- filtering by ISO week and day of year
- zero dependencies

Over uschedule:
- due to supervision, no centralized daemon required
- filtering by ISO week and day of year

Over systemd timers:
- mnemonic syntax
- less complexity for jobs, no need for a `<job>.timer` file distinct
  from the `<job>.service` file,
- can use a file timestamp to ensure minimum waiting time between two
  runs, even across reboots
- very robust with respect to external time changes
- randomized delays
- variable slack
- ad-hoc usage possible, just run the program from command line

## Rosetta stone

* run five minutes after midnight, every day:
  cron: `5 0 * * *`
  snooze: `-M5`
* run at 2:15pm on the first of every month:
  cron: `15 14 1 * *`
  snooze: `-d1 -H14 -M15`
* run at 10 pm on weekdays:
  cron: `0 22 * * 1-5`
  snooze: `-w1-5 -H22`
* run 23 minutes after midnight, 2am, 4am ..., everyday:
  cron: `23 0-23/2 * * *`
  snooze: `-H/2 -M23`
* run every second week:
  snooze: `-W/2`
* run every 10 days:
  snooze: `-D/10`

## Usage:

	snooze [-nv] [-t timefile] [-T timewait] [-R randdelay] [-s slack] [-d mday] [-m mon] [-w wday] [-D yday] [-W yweek] [-H hour] [-M min] [-S sec] COMMAND...

* `-n`: dry-run, print the next 5 times the command would run.
* `-v`: verbose, print scheduled (and rescheduled) times.
* `-t`, `-T`: see below timefiles
* `-R`: add between 0 and RANDDELAY seconds to the start of the scheduled time.
* `-J`: add between 0 and JITTER seconds to scheduled execution time.
* `-s`: commands are executed even if they are SLACK (default: 60) seconds late.

The durations RANDDELAY and JITTER and SLACK and TIMEWAIT are parsed as seconds,
unless a postfix of `m` for minutes, `h` for hours, or `d` for days is used.

The remaining arguments are patterns for the time fields:

* `-d`: day of month
* `-m`: month
* `-w`: weekday (0-7, sunday is 0 and 7)
* `-D`: day of year
* `-W`: ISO week of year (0..53)
* `-H`: hour
* `-M`: minute
* `-S`: second

The following syntax is used for these options:

* exact match: `-d 3`, run on the 3rd
* alternation: `-d 3,10,27`, run on 3rd, 10th, 27th
* range: `-d 1-5`, run on 1st, 2nd, 3rd, 4th, 5th
* star: `-d '*'`, run every day
* repetition: `-d /5`, run on 5th, 10th, 15th, 20th, 25th, 30th day
* shifted repetition: `-d 2/5`, run on 7th, 12th, 17th, 22nd, 27th day

and combinations of those, e.g. `-d 1-10,15/5,28`.

The defaults are `-d* -m* -w* -D* -W* -H0 -M0 -S0`, that is, every midnight.

Note that *all* patterns need to match (contrary to cron where either
day of month *or* day of week matches), so `-w5 -d13` only runs on
Friday the 13th.

## Timefiles

Optionally, you can keep track of runs in time files, using `-t` and
optionally `-T`.

When `-T` is passed, execution will not start earlier than the mtime
of TIMEFILE plus TIMEWAIT seconds.

When `-T` is *not* passed, snooze will start finding the first matching time
starting from the mtime of TIMEFILE, and taking SLACK into account.
(E.g. `-H0 -s 1d -t timefile` will start an instant
execution when timefile has not been touched today, whereas without `-t`
this would always wait until next midnight.)

If TIMEFILE does not exist, it will be assumed outdated enough to
ensure earliest execution.

snooze does not update the timefiles, your job needs to do that!
Only mtime is looked at, so touch(1) is good.

## Exact behavior

* snooze parses the option flags and computes the first time the
  date pattern matches, as a symbolic date
* if a timefile is specified, the time is upped to timefile + timewait seconds
* if a random delay is requested, it is added
* snooze computes how far this event is in the future
* snooze sleeps that long, but at most 5 minutes
* after waking, snooze recomputes how far the event is in the future
* if the event is in the past, but fewer than SLACK seconds, snooze
  execs the command.  You need to ensure (by setting up supervision)
  snooze runs again after that!
* if we woke due to a SIGALRM, the command is executed immediately as well
* if we notice time moved backwards, recompute the time until the event
* if the event is in the future, recompute the time it takes, possibly
  considering shifting of the system time or timezone changes
  (timezone reload only tested on glibc)
* If no command was given, just return with status 0
* and so on...

## Common usages

Run a job like cron, every day at 7am and 7pm:

	exec snooze -H7,19 rdumpfs / /data/dump/mybox 2>&1

Run a job daily, never twice a day:

	exec snooze -H0 -s 1d -t timefile \
		sh -c 'run-parts /etc/cron.daily; touch timefile'

Use snooze inline, run a mirror script every hour at 30 minutes past,
but ensure there are at least 20 minutes in between.

	set -e
	snooze -H'*' -M30 -t timefile -T 20m
	touch timefile  # remove this if instantly retrying on failure were ok
	mirrorallthethings
	touch timefile

Use snooze inline, cron-style mail:

	set -e
	snooze ...
	actualjob >output 2>&1 ||
		mail -s "$(hostname): snooze job failed with status $?" root <output

Snooze for rate-limiting a general purpose runit service: don't
restart faster than every two minutes. (Note that after a crash with a
daemon runtime of more than two minutes, it will be restarted
immediately):

	set -e
	snooze -H'*' -M'*' -S'*' -t timefile -T 2m
	touch timefile
	exec mydaemond

## Installation

Use `make all` to build, `make install` to install relative to `PREFIX`
(`/usr/local` by default).  The `DESTDIR` convention is respected.
You can also just copy the binary into your `PATH`.

## Copyright

snooze is in the public domain.

To the extent possible under law, Leah Neukirchen <leah@vuxu.org>
has waived all copyright and related or neighboring rights to this work.

http://creativecommons.org/publicdomain/zero/1.0/