2 # Copyright (C) 2019 Ian Kelling
3 # SPDX-License-Identifier: AGPL-3.0-or-later
5 # usage: runs once every 15 seconds unless any args are passed, or we
6 # then just runs once and have verbose output. On battery power, run
9 if [ -z "$BASH_VERSION" ]; then echo "error: shell is not bash" >&2; exit 1; fi
11 if [[ $EUID != 1000 ]]; then
12 echo "$0: error, expected to be user 1000"
16 source /a
/bin
/errhandle
/err
17 status_file
=/dev
/shm
/iank-status
23 for p
in ~
/.gem
/ruby
/*/bin
; do
37 p
() { printf "%s\n" "$*"; }
38 # log-once COUNT NAME [MESSAGE]
40 if type -p ifne
&>/dev
/null
; then
41 /usr
/local
/bin
/log-once
"$@" | ifne
mail -s "$HOSTNAME: system-status $2" root@localhost
46 if type -p ifne
&>/dev
/null
; then
47 /usr
/local
/bin
/log-once
"$@" | ifne
mail -s "$HOSTNAME: system-status $2" daylert@iankelling.org
57 # todo, consider migrating some of these alerts into prometheus
59 chars
=("${first_chars[@]}")
61 services
=( epanicclean
)
71 if systemctl show
-p SubState
--value ${services[@]} |
grep -E -v '^(running|)$' &>/dev
/null
; then
72 for s
in ${services[@]}; do
73 if [[ $
(systemctl show
-p SubState
--value $s 2>&1) != running
]]; then
79 p
${bads[*]} | lo
-240 mysers
86 prometheus-node-exporter
87 prometheus-alertmanager
91 if systemctl show
-p SubState
--value ${services[@]} |
grep -E -v '^(running|)$' &>/dev
/null
; then
92 for s
in ${services[@]}; do
93 if [[ $
(systemctl show
-p SubState
--value $s 2>&1) != running
]]; then
99 p
${bads[*]} | lo
-240 prom
104 # this section copied from servicepid()
106 pid
=$
(systemctl show
--property MainPID
--value $unit ||
:)
110 dir
=/sys
/fs
/cgroup
/system.slice
111 if [[ ! -d $dir ]]; then
112 dir
=/sys
/fs
/cgroup
/systemd
/system.slice
114 pid
=$
(head -n1 $dir/${unit%.service}.service
/cgroup.procs ||
:)
117 if [[ ! $pid ]]; then
122 if [[ -e /a
/bin
/bash_unpublished
/source-state
]]; then
123 # /a gets remounted due to btrbk, ignore error code for file doesnt exist
124 source /a
/bin
/bash_unpublished
/source-state ||
[[ $?
== 1 ]]
128 ## check if last snapshot was recent
129 old_snap_limit
=$
(( 3 * 60 * 60 ))
132 o
) btrbk_root
=/mnt
/o
/btrbk
;;
133 *) btrbk_root
=/mnt
/root
/btrbk
;;
135 # this section generally copied from btrbk scripts, but
136 # this part modified to speed things up by about half a second.
137 # I'm not sure if its quite as reliable, but it looks pretty safe.
138 # Profiled it using time and also adding to the top of the file:
140 # PS4='+ $(date "+%2N") '
141 # allow failure in case there are no snapshots yet.
142 # shellcheck disable=SC2012
144 files
=($btrbk_root/$vol.20*)
146 if (( ! ${#files[@]} )); then
149 snaps
=($
(ls -1avdr "${files[@]}" 2>/dev
/null |
head -n1 ||
: ))
152 for s
in ${snaps[@]}; do
154 t
=$
(date -d $
(sed -r 's/(.{4})(..)(.{5})(..)(.*)/\1-\2-\3:\4:\5/' <<<${file#$vol.}) +%s
)
155 if (( t
> maxtime
)); then
160 last_snap_age
=$
(( now
- maxtime
))
161 last_snap_hours
=$
(( last_snap_age
/ 60 / 60 ))
162 if (( last_snap_age
> old_snap_limit
)); then
163 chars
+=(OLD-SNAP-
${last_snap_hours}h
)
164 snapshotmsg
="/$vol snapshot older than 4 hours"
165 if [[ $MAIL_HOST == "$HOSTNAME" ]]; then
166 p
"$snapshotmsg" | lo
-1 old-snapshot
168 # not bothering to get info on all volumes if we find an old one.
174 if [[ $MAIL_HOST == "$HOSTNAME" ]]; then
177 glob
=(/m
/md
/bounces
/new
/*)
178 if [[ -e ${glob[0]} ]]; then
180 bouncemsg
="message in /m/md/bounces/new"
182 p
$bouncemsg | loday
-1 bounce
183 # emails without the S (seen) flag. this only checks the last flag,
184 # but its good enough for me.
185 glob
=(/m
/md
/alerts
/{new
,cur
}/!(*,S
))
186 if [[ -e ${glob[0]} ]]; then
190 glob
=(/m
/md
/daylert
/{new
,cur
}/!(*,S
))
191 if [[ -e ${glob[0]} ]]; then
196 if [[ $
(systemctl is-active btrbk.timer
) != active
]]; then
200 p
"$bbkmsg" | lo
-480 btrbk.timer
204 # commented out, only using timetrap retrospectively.
205 # # clock us out in timetrap if are idle too long
206 # if [[ -e /p/.timetrap.db ]]; then
208 # if type -p xprintidle &>/dev/null && xidle=$(xprintidle 2>/dev/null); then
209 # if [[ $xidle == [0-9]* ]]; then
210 # sheet=$(sqlite3 /p/.timetrap.db "select sheet from entries where end is NULL;")
212 # if [[ $sheet == w ]]; then
215 # if [[ $sheet && $xidle -gt $idle ]]; then
221 else # end if $MAIL_HOST
222 rmg
/home
/iank
/cron-errors
/bounce
* \
223 /home
/iank
/cron-errors
/btrbk.timer
* \
224 /home
/iank
/cron-errors
/old-snapshot
*
227 if ip l show tunfsf
&>/dev
/null
; then
228 # this is for tracking dns over tls issue, which
229 # fixvpndns() in brc2 fixes.
230 stat
=$
(resolvectl dnsovertls tunfsf
2>/dev
/null ||
: )
231 read -r _ _ _ istls
<<<"$stat"
235 printf "%s\n" "$istls" | ts
>> /tmp
/istls.log
241 # We do this once every 5 minutes, since this is not a grave problem.
242 # For formatted elisp, see /b/ds/unsaved-buffers.el
243 elisp
='(format "%s" (-reduce-from (lambda (acc buf) (let ((bpath (buffer-file-name buf))) (if (and bpath (buffer-modified-p buf)) (cons bpath acc ) acc))) nil (buffer-list)))'
244 if [[ ! $last_emacs_check ||
$emacsfiles ]] ||
(( last_emacs_check
< EPOCHSECONDS
- 300 )); then
245 if pgrep
-G iank
-u iank
-f 'emacs --daemon' &>/dev
/null
; then
246 # i dun care if this fails
247 emacsfiles
="$(timeout 1 emacsclient -a /usr/bin/true --eval "$elisp" 2>/dev/null | sed '/^"nil
"$/d;s/^"(/E
: /;s
/)"$//' ||:)"
248 if [[ $emacsfiles ]]; then
249 chars
+=("$emacsfiles")
252 last_emacs_check
=$EPOCHSECONDS
256 glob
=(/nocow
/btrfs-stale
/*)
257 if [[ -e ${glob[0]} ]]; then
261 if [[ $
(find /var
/mail -type f \
! -empty -print -quit) ]]; then
262 var_mail_msg
="message in /var/mail"
264 p
$var_mail_msg | loday
-1 var_mail
266 # early in install process, we dont have permission yet for exiqgrep.
267 # 1100 helps allow for system restarts
268 qlen
=$
(/usr
/sbin
/exiqgrep
-o 1100 -c -b |
awk '{print $1}') ||
:
271 qmsg
="queue length $qlen"
275 # No point in emailing about the mailq on a host where we don't
278 p
$qmsg | loday
-120 qlen
281 rmg
/home
/iank
/cron-errors
/qlen
*
287 # todo: make this robust to the case of /a not being mounted
288 if ! make -C /b
/ds
-q ~
/.local
/distro-begin
2>/dev
/null ||
[[ $
(<~
/.local
/distro-begin
) != 0 ]]; then
293 if ! make -C /b
/ds
-q ~
/.local
/distro-end
2>/dev
/null ||
[[ $
(<~
/.local
/distro-end
) != 0 ]]; then
297 # these conditions are so we dont have an overly verbose prompt
298 if $begin && $end; then
305 source /a
/bin
/ds
/script-files
307 # shellcheck disable=SC2043
312 # the / 60 makes it 0-59 seconds less strict, +1 to help make sure we
313 # dont have any false positives.
314 fmin
=$
(( (fsec
- now
+ 1 ) / 60 ))
315 fminplus
=$
(( fmin
+ 60*24 ))
316 # Filesystem files get copied, so find any newer than the last run.
317 # The rest are hueristics:
318 # Given the last time we added a file in git, is that newer than the last conflink run.
319 # Given new files not added to git, were they modified more recently than the last conflink? but,
320 # push their modification time back by a day so we can develop them before needing to add them to git.
322 all_dirs
=({/a
/bin
/ds
,/p
/c
}{/filesystem
,/machine_specific
/$HOSTNAME/filesystem
})
323 # This part is copied from conflink
324 for x
in /p
/c
/machine_specific
/*.hosts
/a
/bin
/ds
/machine_specific
/*.hosts
; do
325 if grep -qxF $HOSTNAME $x; then all_dirs
+=( ${x%.hosts} ); fi
328 # Just because i forget a lot, -mmin -NUM means files modified <= NUM minutes ago
329 if (( fmin
< 0 )) && [[ $
(find ${all_my_scripts[@]} ${all_dirs[@]} -mmin $fmin -type f
-print -quit 2>/dev
/null
) ]]; then
330 v conflink newer filesystem files
335 for d
in /a
/bin
/distro-setup
/p
/c
; do
336 [[ -d $d ]] ||
continue
338 if [[ ! -e .git
]]; then
339 # some hosts i dont push all of /p/c
342 if (( $
(date -d "$(git log --diff-filter=ACR --format=%aD -1)" +%s
) > fsec
)); then
343 v conflink
: newer files checked
in to git
351 done < <(git ls-files
-o --exclude-standard)
352 if [[ ${untracked[0]} && $
(find "${untracked[@]}" -mmin $fminplus -type f
-print -quit) ]]; then
353 v conflink
: untracked
in $d
361 if [[ ! -e $f || $
(<$f) != 0 ]]; then
362 v conflink
: last run not found or failed
369 # if [[ $(grep -v "exim user lost privilege for using -C option" /var/log/exim4/paniclog 2>/dev/null ||:) ]]; then
370 if [[ -s /var
/log
/exim
4/paniclog
]]; then
372 # leave it up to epanic-clean to send email notification
375 if [[ ! -e $status_file ||
-w $status_file ]]; then
376 if [[ -e /a
/bin
/bash_unpublished
/source-state
]]; then
377 cat /a
/bin
/bash_unpublished
/source-state
>$status_file
380 if [[ ${chars[*]} ]]; then
381 echo "ps_char=\"${chars[*]} \$ps_char\"" >>$status_file
386 # This prevents me having to mute notifications when I'm going to bed.
391 if lock_info
=$
(xscreensaver-command
-time); then
392 if [[ $lock_info != *non-blanked
* ]]; then
398 midnight
=$
(date -d 00:00 +%s
)
399 mdiff
=$
(( EPOCHSECONDS
- midnight
))
400 if $locked && (( mdiff
< 6 || mdiff
> 21 )); then
401 case $
(pactl get-sink-mute @DEFAULT_SINK@ |
awk '{print $2}') in
405 pactl set-sink-mute @DEFAULT_SINK@ true
409 if ! $locked && (( mdiff
> 6 || mdiff
< 12 )) && [[ ! -e /tmp
/ianknap
]]; then
410 case $
(pactl get-sink-mute @DEFAULT_SINK@ |
awk '{print $2}') in
414 pactl set-sink-mute @DEFAULT_SINK@ false
420 # use this if we want to do something just once per minute
432 if [[ -e /sys
/class
/power_supply
/AC
/online
&& $
(</sys
/class
/power_supply
/AC
/online
) == 0 ]]; then
437 if systemctl
-q is-active bitcoind
; then
449 # ensure our long operations are one line so we are not prone errors
450 # from this file being modified.