Skip to content
This repository has been archived by the owner on Dec 18, 2019. It is now read-only.

Commit

Permalink
Merge pull request #139 from austincunningham/fix-sop-links
Browse files Browse the repository at this point in the history
fix(fix sop url) : fix (INTLY-1983)
  • Loading branch information
austincunningham committed Jun 28, 2019
2 parents 40a16d1 + 0319d31 commit f6e717a
Show file tree
Hide file tree
Showing 2 changed files with 10 additions and 9 deletions.
15 changes: 8 additions & 7 deletions deploy/monitor/mss_prometheus_rule.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@ spec:
annotations:
description: "The mobile-security-service has been down for more than 5 minutes. "
summary: "The mobile-security-service is down. For more information see on the MSS at https://github.com/aerogear/mobile-security-service"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-mss.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-mss.md"
- alert: MobileSecurityServiceConsoleDown
expr: absent(kube_endpoint_address_available{endpoint="mobile-security-service-application"} == 1)
for: 5m
Expand All @@ -31,7 +31,7 @@ spec:
annotations:
description: "The mobile-security-service admin console has been down for more than 5 minutes. "
summary: "The mobile-security-service admin console endpoint has been unavailable for more that 5 minutes. For more information see on the MSS at https://github.com/aerogear/mobile-security-service"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-mss.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-mss.md"
- alert: MobileSecurityServiceDatabaseDown
expr: absent(kube_pod_container_status_running{namespace="mobile-security-service",container="database"}==1)
for: 5m
Expand All @@ -40,7 +40,7 @@ spec:
annotations:
description: "The mobile-security-service-db pod has been down for more than 5 minutes"
summary: "The mobile-security-service-db is down. For more information see on the MSS at https://github.com/aerogear/mobile-security-service"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-mss.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-mss.md"
- alert: MobileSecurityServicePodCPUHigh
expr: "(rate(process_cpu_seconds_total{job='mobile-security-service-application'}[1m])) > (((kube_pod_container_resource_limits_cpu_cores{namespace='mobile-security-service',container='application'})/100)*90)"
for: 5m
Expand All @@ -49,7 +49,7 @@ spec:
annotations:
description: "The mobile-security-service pod has been at 90% CPU usage for more than 5 minutes"
summary: "The mobile-security-service is reporting high cpu usage for more that 5 minutes. For more information see on the MSS at https://github.com/aerogear/mobile-security-service"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-mss.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-mss.md"
- alert: MobileSecurityServicePodMemoryHigh
expr: "(process_resident_memory_bytes{job='mobile-security-service-application'}) > (((kube_pod_container_resource_limits_memory_bytes{namespace='mobile-security-service',container='application'})/100)*90)"
for: 5m
Expand All @@ -58,7 +58,7 @@ spec:
annotations:
description: "The mobile-security-service pod has been at 90% memory usage for more than 5 minutes"
summary: "The mobile-security-service is reporting high memory usage for more that 5 minutes. For more information see on the MSS at https://github.com/aerogear/mobile-security-service"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-mss.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-mss.md"
- alert: MobileSecurityServiceApiHighRequestDuration
expr: "api_requests_duration_seconds{job='mobile-security-service-application', quantile='0.5'} > 30"
for: 5m
Expand All @@ -67,7 +67,7 @@ spec:
annotations:
description: "The mobile-security-service api has had http requests latency longer that 30 seconds for more than 5 minutes"
summary: "The mobile-security-service is reporting high request latency for more that 5 minutes. For more information see on the MSS at https://github.com/aerogear/mobile-security-service"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-mss.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-mss.md"
- alert: MobileSecurityServiceApiHighConcurrentRequests
expr: "api_requests_in_flight{job='mobile-security-service-application'} > 50"
for: 5m
Expand All @@ -76,7 +76,7 @@ spec:
annotations:
description: "The mobile-security-service api has had 50 concurrent requests for more than 5 minutes"
summary: "The mobile-security-service is reporting high request concurrency for more that 5 minutes. For more information see on the MSS at https://github.com/aerogear/mobile-security-service"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-mss.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-mss.md"
- alert: MobileSecurityServiceApiHighRequestFailure
expr: "rate(api_requests_failure_total{job='mobile-security-service-application'}[1h])>10"
for: 1h
Expand All @@ -85,4 +85,5 @@ spec:
annotations:
description: "The mobile-security-service api has reported more that 10 request failures in an hour"
summary: "The mobile-security-service is reporting a high request failure over an hour. For more information see on the MSS at https://github.com/aerogear/mobile-security-service"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-mss.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-mss.md"
4 changes: 2 additions & 2 deletions deploy/monitor/prometheus_rule.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -22,12 +22,12 @@ spec:
annotations:
description: "The mobile-security-service-operator has been down for more than 5 minutes. "
summary: "The mobile-security-service-operator is down. For more information see on the MSS operator https://github.com/aerogear/mobile-security-service-operator"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-operator.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-operator.md"
- alert: MobileSecurityServicePodcount
annotations:
description: The Pod count for the mobile-security-service has changed in the last 5 minutes.
summary: Pod count for namespace mobile-security-service is {{ printf "%.0f" $value }}. Expected 3 pods. For more information see on the MSS operator https://github.com/aerogear/mobile-security-service-operator"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/SOP/SOP-operator.md"
sop_url: "https://github.com/aerogear/mobile-security-service-operator/blob/0.1.4/SOP/SOP-operator.md"
expr: |
(1-absent(kube_pod_status_ready{condition="true", namespace="mobile-security-service"})) or sum(kube_pod_status_ready{condition="true", namespace="mobile-security-service"}) != 3
for: 5m
Expand Down

0 comments on commit f6e717a

Please sign in to comment.