andreas/heartbeat
Public Access
1
0
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases 44 Wiki Activity

Fix rtt, including bug in time compute

Browse Source
This commit is contained in:
Andreas Wrede
2026-04-01 19:41:53 -04:00
parent 090d341244
commit 460d2be9e9
13 changed files with 1366 additions and 372 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.hb.yaml
+212 -126
View File
@@ -7,33 +7,125 @@ logfile: "/home/andreas/logs/heartbeat/heartbeat.log"
logfmt: "msg"
grace: 40
interval: 10
watchhosts:
# "localhost":
# "haschloss" :
# "cotgate":
"wentworth":
notify: +4915123456789
src: "signal"
"y":
notify: +4915123456789
src: "signal"
"winter":
notify: +14168226179
src: "signal"
dyndnshosts: {"haschloss", "wayback", "wertvoll", "weekend", "cotgate", "rvgate", "draper", "eris"}
# Notification Channels - Define notification providers centrally
# Each channel has a type (pushover, email, signal, mattermost) and type-specific configuration
notification_channels:
pushover_standard:
type: pushover
token: ac7NLX2rPjXFareeDgLpXNoDf4iFmf
user: uDhH33UjQQDYtNzJb1ThRiWb9ingGK
signal_andreas:
type: signal
cli_path: /usr/local/bin/signal-cli
user: +14168226179
recipient: +14168226179
email_andreas:
type: email
recipients: [aew.hbd.notify@wrede.ca]
sender: aew.hbd@wrede.ca
smtp_server: smtp.fastmail.com
smtp_port: 587
smtp_user: andreas@wrede.ca
smtp_password: pvtvefyp5gbhnch2
# Example additional channels (commented out)
# pushover_urgent:
# type: pushover
# token: your-app-token
# user: your-user-key
#
mattermost_devops:
type: mattermost
host: mattermost.example.com
token: webhook-token
channel: devops-alerts
username: heartbeat-bot
icon: https://example.com/heartbeat-icon.png
# Default notification channels (used if host doesn't specify channels)
default_notification_channels: [pushover_standard]
# Host definitions - combines threshold mapping, watch status, DNS updates, and notifications
hosts:
wentworth:
threshold_config: default
watch: true
notification_channels: [pushover_standard]
dyndns: false
y:
threshold_config: default
watch: true
notification_channels: [pushover_standard]
dyndns: false
winter:
threshold_config: default
watch: true
notification_channels: [pushover_standard]
dyndns: false
wally:
threshold_config: freebsd_server
watch: false
notification_channels: [pushover_standard]
dyndns: false
eris:
threshold_config: truenas_server
watch: false
notification_channels: [pushover_standard]
dyndns: false
haschloss:
threshold_config: default
watch: false
dyndns: true
wayback:
threshold_config: default
watch: false
notification_channels: [pushover_standard]
dyndns: true
wertvoll:
threshold_config: default
watch: false
notification_channels: [pushover_standard]
dyndns: true
weekend:
threshold_config: default
watch: false
notification_channels: [pushover_standard]
dyndns: true
cotgate:
threshold_config: default
watch: false
dyndns: true
rvgate:
threshold_config: default
watch: false
dyndns: true
draper:
threshold_config: default
watch: false
notification_channels: [pushover_standard]
dyndns: true
# Hosts to drop/ignore
drophosts: {"unknown", "wookie15", "wort"}
nsupdate_bin: "/usr/local/bin/nsupdate"
pushover_token: "ac7NLX2rPjXFareeDgLpXNoDf4iFmf"
pushover_user: "uDhH33UjQQDYtNzJb1ThRiWb9ingGK"
pushsrv: "pushover"
dyndomains: {"wrede.org"}
toemail: ["aew.hbd.notify@wrede.ca"]
fromemail: "aew.hbd@wrede.ca"
smtpserver: "smtp.fastmail.com"
smtpuser: "andreas@wrede.ca"
smtppassword: "r8psra6wj6gcakkp"
smtpport: 587
ws_port: 50005
# wss_port: 50006 # Commented out - use plain WebSocket instead of secure WSS
@@ -49,120 +141,114 @@ journal_file: messages.journal # Base filename
journal_max_size: 104857600 # Max size (100MB default)
journal_max_backups: 10 # Number of backups to keep
thresholds:
threshold_configs:
default:
cpu_monitor:
cpu_percent:
warning: 80.0
critical: 90.0
memory_monitor:
percent:
warning: 3.0
critical: 95.0
disk_monitor:
partitions:
/:
percent:
warning: 85.0
critical: 90.0
rtt:
y:
thresholds:
cpu_monitor:
cpu_percent:
warning: 80.0
critical: 90.0
memory_monitor:
percent:
warning: 85.0
critical: 95.0
disk_monitor:
partitions:
/:
percent:
warning: 85.0
critical: 90.0
rtt:
warning: 30
critical: 250.0
freebsd_server:
cpu_monitor:
cpu_percent:
warning: 80.0
critical: 90.0
memory_monitor:
percent:
warning: 3.0
critical: 95.0
disk_monitor:
partitions:
/:
percent:
warning: 85.0
critical: 90.0
nagios_runner:
# overall_status_code:
# warning: 1
# critical: 2
# operator: ">="
load_status:
warning: WARNING
critical: CRITICAL
operator: "=="
UPS_load:
display: "{ups_output}"
warning: 70
critical: 80
operator: ">="
UPS_status_code:
display: "{ups_output}"
warning: 1
critical: 2
operator: ">="
nextcloud_apps_status_code:
display: "{nextcloud_apps_output}"
warning: 1
critical: 2
operator: ">="
rtt:
y:
thresholds:
cpu_monitor:
cpu_percent:
warning: 80.0
critical: 90.0
memory_monitor:
percent:
warning: 3.0
critical: 95.0
disk_monitor:
partitions:
/:
percent:
warning: 85.0
critical: 90.0
nagios_runner:
# overall_status_code:
# warning: 1
# critical: 2
# operator: ">="
load_status:
warning: WARNING
critical: CRITICAL
operator: "=="
UPS_load:
display: "{ups_output}"
warning: 70
critical: 80
operator: ">="
UPS_status_code:
display: "{ups_output}"
warning: 1
critical: 2
operator: ">="
nextcloud_apps_status_code:
display: "{nextcloud_apps_output}"
warning: 1
critical: 2
operator: ">="
rtt:
warning: 30
critical: 250.0
truenas_server:
cpu_monitor:
cpu_percent:
warning: 80.0
critical: 90.0
memory_monitor:
percent:
warning: 3.0
critical: 95.0
disk_monitor:
partitions:
/:
percent:
warning: 85.0
critical: 90.0
nagios_runner:
# overall_status_code:
# warning: 1
# critical: 2
# operator: ">="
load_status:
warning: WARNING
critical: CRITICAL
operator: "=="
UPS_load:
display: "{ups_output}"
warning: 70
critical: 80
operator: ">="
UPS_status_code:
display: "{ups_output}"
warning: 1
critical: 2
operator: ">="
nextcloud_apps_status_code:
display: "{nextcloud_apps_output}"
warning: 1
critical: 2
operator: ">="
rtt:
y:
thresholds:
cpu_monitor:
cpu_percent:
warning: 80.0
critical: 90.0
memory_monitor:
percent:
warning: 3.0
critical: 95.0
disk_monitor:
partitions:
/:
percent:
warning: 85.0
critical: 90.0
nagios_runner:
# overall_status_code:
# warning: 1
# critical: 2
# operator: ">="
load_status:
warning: WARNING
critical: CRITICAL
operator: "=="
UPS_load:
display: "{ups_output}"
warning: 70
critical: 80
operator: ">="
UPS_status_code:
display: "{ups_output}"
warning: 1
critical: 2
operator: ">="
nextcloud_apps_status_code:
display: "{nextcloud_apps_output}"
warning: 1
critical: 2
operator: ">="
rtt:
warning: 30
critical: 250.0
host_threshold_mapping:
# Critical production servers
wally: freebsd_server
eris: truenas_server