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. On battery power, run once per minute.
8 if [ -z "$BASH_VERSION" ]; then echo "error: shell is not bash" >&2; exit 1; fi
10 source /a
/bin
/errhandle
/err
11 status_file
=/dev
/shm
/iank-status
17 for p
in ~
/.gem
/ruby
/*/bin
; do
31 # log-once COUNT NAME [MESSAGE]
33 /usr
/local
/bin
/log-once
"$@" | ifne
mail -s "$HOSTNAME: system-status $2" root@localhost
37 /usr
/local
/bin
/log-once
"$@" | ifne
mail -s "$HOSTNAME: system-status $2" daylerts@iankelling.org
40 # todo, consider migrating some of these alerts into prometheus
42 chars
=("${first_chars[@]}")
51 if systemctl show
-p SubState
--value ${services[@]} |
egrep -v '^(running|)$'; then
52 for s
in ${services[@]}; do
53 if [[ $
(systemctl show
-p SubState
--value $s) != running
]]; then
60 lo
-240 mysers
${bads[*]}
63 prometheus-node-exporter
64 prometheus-alertmanager
70 if systemctl show
-p SubState
--value ${services[@]} |
egrep -v '^(running|)$'; then
71 for s
in ${services[@]}; do
72 if [[ $
(systemctl show
-p SubState
--value $s) != running
]]; then
78 lo
-240 prom
${bads[*]}
82 # clock us out in timetrap if are idle too long
83 if [[ -e /p
/.timetrap.db
]]; then
85 if type -p xprintidle
&>/dev
/null
&& xidle
=$
(xprintidle
2>/dev
/null
); then
86 if [[ $xidle == [0-9]* ]]; then
87 sheet
=$
(sqlite3
/p
/.timetrap.db
"select sheet from entries where end is NULL;")
89 if [[ $sheet == w
]]; then
92 if [[ $sheet && $xidle -gt $idle ]]; then
100 if ip l show tunfsf
&>/dev
/null
; then
101 # this is for tracking dns over tls issue, which
102 # fixvpndns() in brc2 fixes.
103 stat
=$
(resolvectl dnsovertls tunfsf
2>/dev
/null ||
: )
104 read _ _ _ istls
<<<"$stat"
108 printf "%s\n" "$istls" | ts
>> /tmp
/istls.log
115 if pgrep
-G iank
-u iank
-f 'emacs --daemon' &>/dev
/null
; then
116 emacsfiles
="$(emacsclient --eval "$
(cat /usr
/local
/bin
/unsaved-buffers.el
)"| sed '/^"nil
"$/d;s/^"(/E
: /;s
/)"$//')"
117 if [[ $emacsfiles ]]; then
118 chars
+=("$emacsfiles")
122 glob
=(/nocow
/btrfs-stale
/*)
123 if [[ -e ${glob[0]} ]]; then
127 if [[ $
(find /var
/mail -type f \
! -empty -print -quit) ]]; then
128 var_mail_msg
="message in /var/mail"
130 loday
-1 var_mail
$var_mail_msg
133 glob
=(/m
/md
/bounces
/new
/*)
134 if [[ -e ${glob[0]} ]]; then
136 bouncemsg
="message in /m/md/bounces/new"
138 loday
-1 bounce
$bouncemsg
139 # emails without the S (seen) flag. this only checks the last flag,
140 # but its good enough for me.
141 glob
=(/m
/md
/alerts
/{new
,cur
}/!(*,S
))
142 if [[ -e ${glob[0]} ]]; then
146 glob
=(/m
/md
/daylerts
/{new
,cur
}/!(*,S
))
147 if [[ -e ${glob[0]} ]]; then
152 tmp
=(/var
/local
/cron-errors
/mailtest-check
*)
153 if (( ${#tmp[@]} )); then
156 tmp
=(/var
/local
/cron-errors
/mailtest-slow
*)
157 if (( ${#tmp[@]} )); then
161 # early in install process, we dont have permission yet for exiqgrep.
162 # 1100 helps allow for system restarts
163 qlen
=$
(/usr
/sbin
/exiqgrep
-o 1100 -c -b |
awk '{print $1}') ||
:
166 qmsg
="queue length $qlen"
170 # No point in emailing about the mailq on a host where we don't
173 loday
-120 qlen
$qmsg
178 if ! make -C /b
/ds
-q ~
/.local
/distro-begin ||
[[ $
(<~
/.local
/distro-begin
) != 0 ]]; then
183 if ! make -C /b
/ds
-q ~
/.local
/distro-end ||
[[ $
(<~
/.local
/distro-end
) != 0 ]]; then
187 # these conditions are so we dont have an overly verbose prompt
188 if $begin && $end; then
196 # shellcheck disable=SC2043
201 # the / 60 makes it 0-59 seconds less strict, +1 to help make sure we
202 # dont have any false positives.
203 fmin
=$
(( (fsec
- now
+ 1 ) / 60 ))
204 fminplus
=$
(( fmin
+ 60*24 ))
205 # Filesystem files get copied, so find any newer than the last run.
206 # The rest are hueristics:
207 # Given the last time we added a file in git, is that newer than the last conflink run.
208 # Given new files not added to git, were they modified more recently than the last conflink? but,
209 # push their modification time back by a day so we can develop them before needing to add them to git.
211 all_dirs
=({/a
/bin
/ds
,/p
/c
}{/filesystem
,/machine_specific
/$HOSTNAME/filesystem
})
212 # This part is copied from conflink
213 for x
in /p
/c
/machine_specific
/*.hosts
/a
/bin
/ds
/machine_specific
/*.hosts
; do
214 if grep -qxF $HOSTNAME $x; then all_dirs
+=( ${x%.hosts} ); fi
217 # Just because i forget a lot, -mmin -NUM means files modified <= NUM minutes ago
218 if (( fmin
< 0 )) && [[ $
(find ${all_dirs[@]} -mmin $fmin -type f
-print -quit 2>/dev
/null
) ]]; then
219 v conflink newer filesystem files
224 for d
in /a
/bin
/distro-setup
/p
/c
; do
225 [[ -d $d ]] ||
continue
227 if [[ ! -e .git
]]; then
228 # some hosts i dont push all of /p/c
231 if (( $
(date -d "$(git log --diff-filter=ACR --format=%aD -1)" +%s
) > fsec
)); then
232 v conflink
: newer files checked
in to git
240 done < <(git ls-files
-o --exclude-standard)
241 if [[ ${untracked[0]} && $
(find "${untracked[@]}" -mmin $fminplus -type f
-print -quit) ]]; then
242 v conflink
: untracked
in $d
250 if [[ ! -e $f || $
(<$f) != 0 ]]; then
251 v conflink
: last run not found or failed
258 # if [[ $(grep -v "exim user lost privilege for using -C option" /var/log/exim4/paniclog 2>/dev/null ||:) ]]; then
259 if [[ -s /var
/log
/exim
4/paniclog
]]; then
261 # leave it up to epanic-clean to send email notification
264 source /a
/bin
/bash_unpublished
/source-state
265 if [[ $MAIL_HOST == "$HOSTNAME" ]]; then
267 if [[ $
(systemctl is-active btrbk.timer
) != active
]]; then
271 lo
-480 btrbk.timer
$bbkmsg
273 ## check if last snapshot was within an hour
275 # this section generally copied from btrbk scripts, but
276 # this part modified to speed things up by about half a second.
277 # I'm not sure if its quite as reliable, but it looks pretty safe.
278 # Profiled it using time and also adding to the top of the file:
280 # PS4='+ $(date "+%2N") '
281 # allow failure in case there are no snapshots yet.
282 # shellcheck disable=SC2012
284 files
=(/mnt
/root
/btrbk
/$vol.20*)
287 if (( ${#files[@]} )); then
288 snaps
=($
(ls -1avdr "${files[@]}" 2>/dev
/null |
head -n1 ||
: ))
292 for s
in ${snaps[@]}; do
294 t
=$
(date -d $
(sed -r 's/(.{4})(..)(.{5})(..)(.*)/\1-\2-\3:\4:\5/' <<<${file#$vol.}) +%s
)
295 if (( t
> maxtime
)); then
300 if (( maxtime
< now
- 4*60*60 )); then
302 snapshotmsg
="/o snapshot older than 4 hours"
304 lo
-1 old-snapshot
$snapshotmsg
307 cat /a
/bin
/bash_unpublished
/source-state
>$status_file
309 if [[ ${chars[*]} ]]; then
310 echo "ps_char=\"${chars[*]} \$ps_char\"" >>$status_file
314 # use this if we want to do something just once per minute
327 if [[ -e /sys
/class
/power_supply
/AC
/online
&& $
(</sys
/class
/power_supply
/AC
/online
) == 0 ]]; then
340 # ensure our long operations are one line so we are not prone errors
341 # from this file being modified.