trey/psi-alerts
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Refactored to take into account ssh-agent, and time of last alert

Browse Source
This commit is contained in:
Trey Blancher
2023-08-24 09:59:45 -04:00
parent 15f6697dad
commit 0540fed30b
1 changed files with 85 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

114
psi-alerts.sh
View File

@@ -52,9 +52,25 @@ ssh_user="${SSH_USER}"
ssh_id_path="${SSH_ID_PATH}"
clear_threshold="${CLEAR_THRESHOLD}"
notification_cmd="${NOTIFICATION_CMD}"
notification_hist_cmd="${NOTIFICATION_HIST_CMD}"
notification_opts="${NOTIFICATION_OPTS}"
id_idx="${NOTIFICATION_IDX}"
get_ssh_agent () {
for dir in /tmp/ssh-*; do
if [[ -O ${dir} ]]; then
# only choose the last agent
export SSH_AUTH_SOCK=$(ls ${dir}/agent.* | tail -1)
fi
done
if [[ -S ${SSH_AUTH_SOCK} ]]; then
# we found an ssh_agent socket
true
else
false
fi
}
print_psi () {
local psi_file="${1}"
cat "${(P)$(tr '[[:upper:]]' '[[:lower:]]' <<< "${psi_file}")}"
@@ -101,13 +117,27 @@ send_notice () {
esac
integer notification_id
if ! notification_id=$(ssh -q -i "${ssh_id_path}" "${ssh_user}@${ssh_host}" -p ${ssh_port} \
"${notification_cmd} ${notification_opts} '${host}: PSI ${psi_type} triggered!' '${psi}'"); then
print "Connection to notification daemon failed!" >&2
false
if get_ssh_agent && [[ -S ${SSH_AUTH_SOCK} ]]; then
if ! notification_id=$(ssh -q "${ssh_user}@${ssh_host}" -p ${ssh_port} \
"${notification_cmd} ${notification_opts} '${host}: PSI ${psi_type} triggered!' '${psi}'"); then
print "Connection to notification daemon failed!" >&2
false
else
echo ${notification_id}
true
fi
elif [[ -n "${ssh_id_path}" ]]; then
if ! notification_id=$(ssh -q -i "${ssh_id_path}" "${ssh_user}@${ssh_host}" -p ${ssh_port} \
"${notification_cmd} ${notification_opts} '${host}: PSI ${psi_type} triggered!' '${psi}'"); then
print "Connection to notification daemon failed!" >&2
false
else
echo ${notification_id}
true
fi
else
echo ${notification_id}
true
echo "No SSH notifications configured. Returning." >&2
false
fi
#set +x
}
@@ -171,11 +201,11 @@ is_clear () {
local psi_type="${1}"
local psi_file="$(tr '[[:upper:]]' '[[:lower:]]' <<< "${psi_type}")"
local avg10=$(grep some "${(P)psi_file}" | awk '{print $2}' | awk -F= '{print $2}')
local avg300=$(grep some "${(P)psi_file}" | awk '{print $4}' | awk -F= '{print $2}')
local prev=$(grep some "${(P)psi_file}" | awk '{print $5}' | awk -F= '{print $2}')
if [[ ${avg10} -lt ${clear_threshold} ]]; then
if [[ ${avg300} -lt ${clear_threshold} ]]; then
true
else
false
@@ -213,35 +243,69 @@ check_dunst_id_is_visible () {
local dunst_id="${1}"
typeset -a ids
if ids=$(ssh -q "${ssh_host}" -p ${ssh_port} \
"dunstctl history | jq '.data[0][][${id_idx}].data'"); then
echo "Connection to dunst failed!" >&2
return 2
if ! ids=$(ssh -q "${ssh_host}" -p ${ssh_port} -l "${ssh_user}" \
"${notification_hist_cmd} | jq '.data[0][].id.data'"); then
if ! ids=$(ssh -qi "${ssh_id_path}" -p ${ssh_port} -l "${ssh_user}" \
"${ssh_host}" "${notification_hist_cmd} | jq '.data[0][].id.data'"); then
echo "Connection to dunst failed!" >&2
return 2
fi
fi
# if the alert is visible, it's not in the dunst history
if grep -qP "\b${dunst_id}\b" <<< "${ids}"; then
true
else
false
else
true
fi
}
#set -x
local current_alarm=""
local last_alarm=""
typeset -A notice_sent
typeset -A secs
integer last_dunst_id=-1
local last_line=""
journalctl -b 0 -fn 3 -u "${svc}" | \
while read line; do
set -x
while true; do
local line=$(journalctl -u ${svc} -n1)
if [[ "${last_line}" == "${line}" ]]; then
# line hasn't changed since last run, do nothing
sleep 1
continue
fi
last_line="${line}"
local now=$(date +%s)
local last_timestamp=$(date -d $(awk '{print $1" "$2" "$3}' <<< "${line}") +%s)
local time_diff=$(( now - last_timestamp ))
if (( time_diff >= 3 )); then
# haven't seen a monitor alert for 3 seconds, see if we can clear them
if [[ -n "${current_alarms}" ]]; then
typeset -a alarms=( $(tr '|' ' ' <<< "$current_alarms") )
for alarm in ${alarms}; do
integer elapsed=$(( now - ${secs[${alarm}]} ))
if is_clear "${alarm}" && (( elapsed >= 300 )); then
current_alarms=$(sed -E "s/${alarm}\|?//" <<< "${current_alarms}")
last_alarm=$(awk -F'|' '{print $NF}' <<< "${current_alarms}")
unset "notice_sent[${alarm}]"
unset "secs[${alarm}]"
fi
done
fi
sleep 1
continue
fi
local psi_type="$(grep -Eo "(CPU|MEM|IO) PSI event" <<< "${line}" | grep -Eo "CPU|MEM|IO")"
if [[ -n "${psi_type}" ]]; then
secs+=(${psi_type} $(date +%s))
secs+=(${psi_type} ${now})
if [[ "${psi_type}" != "${last_alarm}" ]]; then
if [[ ! ${notice_sent[${psi_type}]} ]]; then
last_dunst_id=$(exec_notices "${psi_type}" "${current_alarms}")
notice_sent+=(${psi_type} true)
elif (( last_dunst_id >= 0 )) && ! check_dunst_id_is_visible "${last_dunst_id}"; then
elif (( last_dunst_id >= 0 )) && check_dunst_id_is_visible "${last_dunst_id}"; then
last_alarm="${psi_type}"
sleep 1
continue
fi
fi
@@ -253,16 +317,8 @@ while read line; do
current_alarms="${current_alarms}|${psi_type}"
fi
fi
else
typeset -a alarms=( $(tr '|' ' ' <<< "$current_alarms") )
for alarm in ${alarms}; do
integer elapsed=$(( $(date +%s) - ${secs[${alarm}]} ))
if is_clear "${alarm}" && (( elapsed > 300 )); then
current_alarms=$(sed -E "s/${alarm}\|?//" <<< "${current_alarms}")
last_alarm=$(awk -F'|' '{print $NF}' <<< "${current_alarms}")
fi
done
fi
sleep 1
done
#set +x
set +x