Add redesign note to decouple sleep from sensor
[khatus.git] / README.md
CommitLineData
756b9d5a
SK
1khatus
2======
a9038fad 3![mascot](mascot.jpg)
756b9d5a 4
44fc2f3d 5Experimental system-monitor and status (bar) reporter I use with
61c33dc2 6[dwm](https://dwm.suckless.org/) on GNU/Linux.
756b9d5a
SK
7
8![screenshot](screenshot.jpg)
55407653 9
44fc2f3d
SK
10Usage
11-----
12
13In my `~/.xinitrc` I have something like the following:
14
15```sh
16( $BIN/khatus \
17 --wifi_interface 'wlp3s0' \
18| stdbuf -o L tee \
19 >(stdbuf -o L "$BIN"/khatus_bar \
20 -v Opt_Mpd_Song_Max_Chars=10 \
21 -v Opt_Net_Interfaces_To_Show=wlp3s0 \
22 -v Opt_Pulseaudio_Sink=0 \
23 | "$BIN"/khatus_actuate_status_bar_to_xsetroot_name \
24 ) \
25 >(stdbuf -o L "$BIN"/khatus_monitor_energy \
26 | "$BIN"/khatus_actuate_alert_to_notify_send \
27 ) \
28 >(stdbuf -o L "$BIN"/khatus_monitor_errors \
29 | "$BIN"/khatus_actuate_alert_to_notify_send \
30 ) \
31) \
322> >($BIN/twrap.sh >> $HOME/var/log/khatus/main.log) \
331> /dev/null \
34&
35```
36(where `twrap` is a simple script which prefixes a timestamp to each line)
37
38The idea is to support appending any number of ad-hoc, experimental monitors by
39giving maximum flexibility for what to do with the sensor outputs, while
40maintaining some uniformity of msg formats (again, to ease ad-hoc combinations
41(e.g. Does the CPU get hotter when MPD is playing Wu-Tang?)). `khatus_bar`,
42`khatus_monitor_energy` and `khatus_monitor_errors` are just some initial
43examples.
55407653
SK
44
45Design
46------
47
44fc2f3d
SK
48### 2.0
49
50In an effort to simplify the components and their interfaces, I removed the
51concept of a global controller from the previous design (which, at least for
52now, is superfluous), so now it is essentially a pub-sub - parallel publishers
53(sensors) write to a pipe, which is then copied to any number of interested
54subscribers that can filter-out what they need and then do whatever they want
55with the data. Status bar is one such subscriber:
56
8acd36e8 57`P1 > pipe&; P2 > pipe&; ... PN > pipe&; tail -f pipe | tee >(S1) >(S2) ... >(SN) > /dev/null`
44fc2f3d
SK
58
59The cool thing is that, because the pipe is always read (`tail -f ... > /dev/null`),
09caa63e 60the publishers are never blocked, so we get a live stream of events to which we
44fc2f3d
SK
61can attach any number of interested subscribers (` ... tee ... `) and, because
62the pipe is named, if a subscriber needs to - it too can publish something to
63the pipe without being blocked.
64
65```
66parallel +----------+ +----------+ +----------+
67stateless | sensor_1 | | sensor_2 | ... | sensor_n |
68collectors +----------+ +----------+ +----------+
69 | | | |
70 data data data data
71 | | | |
72 V V V V
73multiplexing +-------------+-----------+---------+
74to a pipe |
75 |
76 V
77copying to +-------------+-+---------+---------+
78subscribers | | | |
79 V V V V
80 +------------+ ... +----------------+
81any number of | status bar | | energy monitor |
82parallel +------------+ +----------------+
83subscribers | |
84 V V
85 +----------------+ +-------------+
86 | xsetroot -name | | notify-send |
87 +----------------+ +-------------+
88```
89
90### 1.0
91
92This was an improvement of having everything in one script, but the controller
93was still way too complicated for no good reason.
94
55407653 95```
43e49903
SK
96parallel +----------+ +----------+ +----------+
97stateless | sensor_1 | | sensor_2 | ... | sensor_n |
98collectors +----------+ +----------+ +----------+
99 | | | |
100 data data data data
101 | | | |
102 V V V V
103serial +----------------------------------------------+
104stateful | controller |
105observer +----------------------------------------------+
106 |
107 decision messages
108decision |
109messages |
110copied to |
111any number |
112of interested |
113filter/actuator |
114combinations |
115 |
116 V
117 +-------------+-+---------+---------+
118 | | | |
119 V V V V
120parallel +------------+ +------------+ +------------+
121stateless | filter_1 | | filter_2 | ... | filter_n |
122filters +------------+ +------------+ +------------+
123 | | | |
124 V V V V
125parallel +------------+ +------------+ +------------+
126stateless | actuator_1 | | actuator_2 | ... | actuator_n |
127executors +------------+ +------------+ +------------+
128 | | | |
129 commands commands commands commands
130 | | | |
131 V V V V
132 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
133 ~~~~~~~~~~~~~ operating system ~~~~~~~~~~~~~~~~~
134 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
55407653 135```
7daecd24 136
44fc2f3d
SK
137### 0.x
138
139A single script, re-executed in a loop at some intervals, serially grabbing all
140the needed data and outputting a status bar string, then passed to `xsetroot -name`,
141while saving state in files (e.g. previous totals, to be converted to deltas).
142
143This actually worked surprisingly-OK, but had limitations:
144
145- I use an SSD and want to minimize disk writes
146- not flexible-enough to support my main goal - easy experimentation with
147 various ad-hoc monitors:
148 - I want to set different update intervals for different data sources
149 - I don't want long-running data collectors to block the main loop
150
ec80b440 151### Actuator
43e49903
SK
152Actuator is anything that takes action upon controller messages. A few generic
153ones are included:
154
155- `khatus_actuate_alert_to_notify_send`
156- `khatus_actuate_status_bar_to_xsetroot_name`
157
44fc2f3d
SK
158and, by default, are left disconnected from the data feed, so if desired - need
159to be manually attached when starting `khatus`. See usage section.
ec80b440
SK
160
161### Errors
7daecd24
SK
162Any errors encountered by any sensor are propagated as alerts by the
163controller, which are in turn actualized as desktop notifications by the
43e49903 164`khatus_actuate_alert_to_notify_send` actuator:
ec80b440 165
7daecd24 166![screenshot-self-error-propagation](screenshot-self-error-propagation.jpg)
1341dc96
SK
167
168TODO
169----
170
44fc2f3d 171- status bar templating language
1341dc96
SK
172- retry/cache for sensors fetching flaky remote resources (such as weather)
173- throttling of broken sensors (constantly returns errors)
174- alert specification language
175 - trigger threshold
176 - above/bellow/equal to threshold value
177 - priority
178 - snooze time (if already alerted, when to re-alert?)
179 - text: subject/body
7917538c
SK
180
181Redesign notes
182--------------
183
184- controller should not do formatting
185- need in-memory db for diskless feedback/throttling and cache
91ad8a2f
SK
186- decouple sensor execution from sleep, i.e. a sensor is blocked not by sleep
187 process directly, but by reading of a pipe, to where a sleep process will
188 write a message announcing interval completion and thus signaling execution.
189 This will allow us to manually signal a sensor to update (concretely - I just
190 openned my laptop from sleep and want to force the weather to update
191 immediately); likewise, the sleep process should be blocked on pipe-read
192 until sensor execution is complete - this will allow us to reconfigure
193 intervals at runtime (which seems like a better idea than the above in-memory
194 DB one).
This page took 0.03215 seconds and 4 git commands to generate.