summaryrefslogtreecommitdiffstats
path: root/health/health.d/boinc.conf
diff options
context:
space:
mode:
Diffstat (limited to 'health/health.d/boinc.conf')
-rw-r--r--health/health.d/boinc.conf70
1 files changed, 70 insertions, 0 deletions
diff --git a/health/health.d/boinc.conf b/health/health.d/boinc.conf
new file mode 100644
index 00000000..092a5684
--- /dev/null
+++ b/health/health.d/boinc.conf
@@ -0,0 +1,70 @@
+# Alarms for various BOINC issues.
+
+# Warn on any compute errors encountered.
+ template: boinc_compute_errors
+ on: boinc.states
+ class: Errors
+ type: Computing
+component: BOINC
+ os: *
+ hosts: *
+ lookup: average -10m unaligned of comperror
+ units: tasks
+ every: 1m
+ warn: $this > 0
+ delay: up 1m down 5m multiplier 1.5 max 1h
+ summary: BOINC compute errors
+ info: Average number of compute errors over the last 10 minutes
+ to: sysadmin
+
+# Warn on lots of upload errors
+ template: boinc_upload_errors
+ on: boinc.states
+ class: Errors
+ type: Computing
+component: BOINC
+ os: *
+ hosts: *
+ lookup: average -10m unaligned of upload_failed
+ units: tasks
+ every: 1m
+ warn: $this > 0
+ delay: up 1m down 5m multiplier 1.5 max 1h
+ summary: BOINC failed uploads
+ info: Average number of failed uploads over the last 10 minutes
+ to: sysadmin
+
+# Warn on the task queue being empty
+ template: boinc_total_tasks
+ on: boinc.tasks
+ class: Utilization
+ type: Computing
+component: BOINC
+ os: *
+ hosts: *
+ lookup: average -10m unaligned of total
+ units: tasks
+ every: 1m
+ warn: $this < 1
+ delay: up 5m down 10m multiplier 1.5 max 1h
+ summary: BOINC total tasks
+ info: Average number of total tasks over the last 10 minutes
+ to: sysadmin
+
+# Warn on no active tasks with a non-empty queue
+ template: boinc_active_tasks
+ on: boinc.tasks
+ class: Utilization
+ type: Computing
+component: BOINC
+ os: *
+ hosts: *
+ lookup: average -10m unaligned of active
+ calc: ($boinc_total_tasks >= 1) ? ($this) : (inf)
+ units: tasks
+ every: 1m
+ warn: $this < 1
+ delay: up 5m down 10m multiplier 1.5 max 1h
+ summary: BOINC active tasks
+ info: Average number of active tasks over the last 10 minutes
+ to: sysadmin