diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2021-05-19 12:33:27 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2021-05-19 12:33:27 +0000 |
commit | 841395dd16f470e3c051a0a4fff5b91efc983c30 (patch) | |
tree | 4115f6eedcddda75067130b80acaff9e51612f49 /health/health.d/boinc.conf | |
parent | Adding upstream version 1.30.1. (diff) | |
download | netdata-841395dd16f470e3c051a0a4fff5b91efc983c30.tar.xz netdata-841395dd16f470e3c051a0a4fff5b91efc983c30.zip |
Adding upstream version 1.31.0.upstream/1.31.0
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'health/health.d/boinc.conf')
-rw-r--r-- | health/health.d/boinc.conf | 118 |
1 files changed, 65 insertions, 53 deletions
diff --git a/health/health.d/boinc.conf b/health/health.d/boinc.conf index 25b7f1994..8604abee9 100644 --- a/health/health.d/boinc.conf +++ b/health/health.d/boinc.conf @@ -1,62 +1,74 @@ # Alarms for various BOINC issues. # Warn on any compute errors encountered. -template: boinc_compute_errors - on: boinc.states - os: * - hosts: * -families: * - lookup: average -10m unaligned of comperror - units: tasks - every: 1m - warn: $this > 0 - crit: $this > 1 - delay: up 1m down 5m multiplier 1.5 max 1h - info: average number of compute errors over the last 10 minutes - to: sysadmin + template: boinc_compute_errors + on: boinc.states + class: Computing +component: BOINC + type: Errors + os: * + hosts: * + families: * + lookup: average -10m unaligned of comperror + units: tasks + every: 1m + warn: $this > 0 + crit: $this > 1 + delay: up 1m down 5m multiplier 1.5 max 1h + 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 - os: * - hosts: * -families: * - lookup: average -10m unaligned of upload_failed - units: tasks - every: 1m - warn: $this > 0 - crit: $this > 1 - delay: up 1m down 5m multiplier 1.5 max 1h - info: average number of failed uploads over the last 10 minutes - to: sysadmin + template: boinc_upload_errors + on: boinc.states + class: Computing +component: BOINC + type: Errors + os: * + hosts: * + families: * + lookup: average -10m unaligned of upload_failed + units: tasks + every: 1m + warn: $this > 0 + crit: $this > 1 + delay: up 1m down 5m multiplier 1.5 max 1h + 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 - os: * - hosts: * -families: * - lookup: average -10m unaligned of total - units: tasks - every: 1m - warn: $this < 1 - crit: $this < 0.1 - delay: up 5m down 10m multiplier 1.5 max 1h - info: average number of total tasks over the last 10 minutes - to: sysadmin + template: boinc_total_tasks + on: boinc.tasks + class: Computing +component: BOINC + type: Utilization + os: * + hosts: * + families: * + lookup: average -10m unaligned of total + units: tasks + every: 1m + warn: $this < 1 + crit: $this < 0.1 + delay: up 5m down 10m multiplier 1.5 max 1h + info: average number of total tasks over the last 10 minutes + to: sysadmin # Warn on no active tasks with a non-empty queue -template: boinc_active_tasks - on: boinc.tasks - os: * - hosts: * -families: * - lookup: average -10m unaligned of active - calc: ($boinc_total_tasks >= 1) ? ($this) : (inf) - units: tasks - every: 1m - warn: $this < 1 - crit: $this < 0.1 - delay: up 5m down 10m multiplier 1.5 max 1h - info: average number of active tasks over the last 10 minutes - to: sysadmin + template: boinc_active_tasks + on: boinc.tasks + class: Computing +component: BOINC + type: Utilization + os: * + hosts: * + families: * + lookup: average -10m unaligned of active + calc: ($boinc_total_tasks >= 1) ? ($this) : (inf) + units: tasks + every: 1m + warn: $this < 1 + crit: $this < 0.1 + delay: up 5m down 10m multiplier 1.5 max 1h + info: average number of active tasks over the last 10 minutes + to: sysadmin |