-
Notifications
You must be signed in to change notification settings - Fork 17
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #59 from PhuNguyenMTT/couchbase_20012022
LGTM
- Loading branch information
Showing
9 changed files
with
729 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,19 @@ | ||
| # sumologic-couchbase-monitors | ||
|
|
||
| This script configures Sumo Logic Monitors for Couchbase using Terraform modules. | ||
|
|
||
| This Terraform script configures pre-packaged Sumo Logic Monitors for Couchbase using Terraform modules. | ||
| For installation and configuration, please look at Sumo Logic couchbase [Help Document](https://help.sumologic.com/07Sumo-Logic-Apps/24Web_Servers/Couchbase). | ||
| ## License | ||
|
|
||
| The Couchbase Terraform script is licensed under the apache v2.0 license. | ||
|
|
||
| ## Issues | ||
|
|
||
| Raise issues at [Issues](https://github.com/SumoLogic/terraform-sumologic-sumo-logic-monitor/issues) | ||
|
|
||
| ## Contributing | ||
|
|
||
| * Fork the project on [Github](https://github.com/SumoLogic/terraform-sumologic-sumo-logic-monitor). | ||
| * Make your feature addition or fix bug, write tests and commit. | ||
| * Create a pull request with one of the maintainer as Reviewer. |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,13 @@ | ||
| # Sumo Logic | ||
| # Please replace <YOUR SUMO ACCESS ID> (including brackets) with your Sumo Access ID. https://help.sumologic.com/Manage/Security/Access-Keys | ||
| access_id = "<YOUR SUMO ACCESS ID>" | ||
| # Please replace <YOUR SUMO ACCESS KEY> (including brackets) with your Sumo Access KEY. | ||
| access_key = "<YOUR SUMO ACCESS KEY>" | ||
| # Please update with your deployment, refer: https://help.sumologic.com/APIs/General-API-Information/Sumo-Logic-Endpoints-and-Firewall-Security | ||
| environment = "<DEPLOYMENT>" | ||
| # This flag determines whether to enable all monitors or not. | ||
| monitors_disabled = true | ||
| # The Sumo Logic monitors will be installed in a folder specified by this value. | ||
| folder = "Couchbase" | ||
| # Sumo Logic Couchbase Clusters Filter. For eg: db_cluster=couchbase.prod.01 | ||
| couchbase_data_source = "" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,281 @@ | ||
| { | ||
| "name": "Couchbase", | ||
| "description": "Folder for Couchbase Monitors", | ||
| "type": "MonitorsLibraryFolderExport", | ||
| "children": [ | ||
| { | ||
| "name": "Couchbase - High Memory Usage", | ||
| "description": "This alert fires when memory usage on a node in a Couchbase cluster is high.", | ||
| "type": "MonitorsLibraryMonitorExport", | ||
| "monitorType": "Metrics", | ||
| "evaluationDelay": "0m", | ||
| "alertName": null, | ||
| "runAs": null, | ||
| "notificationGroupFields": [], | ||
| "queries": [ | ||
| { | ||
| "rowId": "A", | ||
| "query": "metric=couchbase_node_memory_free db_cluster=* host=* | avg by db_cluster,host" | ||
| }, | ||
| { | ||
| "rowId": "B", | ||
| "query": "metric=couchbase_node_memory_total db_cluster=* host=* | avg by db_cluster,host" | ||
| }, | ||
| { | ||
| "rowId": "C", | ||
| "query": "(1-#A/#B)*100 along db_cluster,host" | ||
| } | ||
| ], | ||
| "triggers": [ | ||
| { | ||
| "detectionMethod": "MetricsStaticCondition", | ||
| "triggerType": "Critical", | ||
| "timeRange": "-5m", | ||
| "threshold": 80, | ||
| "thresholdType": "GreaterThanOrEqual", | ||
| "occurrenceType": "Always" | ||
| }, | ||
| { | ||
| "detectionMethod": "MetricsStaticCondition", | ||
| "triggerType": "ResolvedCritical", | ||
| "timeRange": "-5m", | ||
| "threshold": 80, | ||
| "thresholdType": "LessThan", | ||
| "occurrenceType": "Always" | ||
| } | ||
| ], | ||
| "notifications": [], | ||
| "isDisabled": true, | ||
| "groupNotifications": true, | ||
| "playbook": "" | ||
| }, | ||
| { | ||
| "name": "Couchbase - Too Many Login Failures", | ||
| "description": "This alert fires when there are too many login failures to a node in a Couchbase cluster.", | ||
| "type": "MonitorsLibraryMonitorExport", | ||
| "monitorType": "Logs", | ||
| "evaluationDelay": "0m", | ||
| "alertName": null, | ||
| "runAs": null, | ||
| "notificationGroupFields": [], | ||
| "queries": [ | ||
| { | ||
| "rowId": "A", | ||
| "query": "db_cluster=* db_system=\"couchbase\" \"login failure\"\n| json \"log\" as _rawlog nodrop\n| if(isEmpty(_rawlog),_raw,_rawlog) as _raw\n| json \"name\" as event_name\n| where event_name=\"login failure\"\n| json \"remote.ip\" as client_ip \n| json \"local.ip\" as couchbase_server\n| count as count by db_cluster, couchbase_server" | ||
| } | ||
| ], | ||
| "triggers": [ | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "Critical", | ||
| "timeRange": "-5m", | ||
| "threshold": 1000, | ||
| "thresholdType": "GreaterThanOrEqual", | ||
| "field": "count" | ||
| }, | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "ResolvedCritical", | ||
| "timeRange": "-5m", | ||
| "threshold": 1000, | ||
| "thresholdType": "LessThan", | ||
| "field": "count" | ||
| } | ||
| ], | ||
| "notifications": [], | ||
| "isDisabled": true, | ||
| "groupNotifications": true, | ||
| "playbook": "" | ||
| }, | ||
| { | ||
| "name": "Couchbase - High CPU Usage", | ||
| "description": "This alert fires when CPU usage on a node in a Couchbase cluster is high.", | ||
| "type": "MonitorsLibraryMonitorExport", | ||
| "monitorType": "Metrics", | ||
| "evaluationDelay": "0m", | ||
| "alertName": null, | ||
| "runAs": null, | ||
| "notificationGroupFields": [], | ||
| "queries": [ | ||
| { | ||
| "rowId": "A", | ||
| "query": "metric=couchbase_bucket_cpu_utilization_rate db_cluster=* host=* | avg by db_cluster,host" | ||
| } | ||
| ], | ||
| "triggers": [ | ||
| { | ||
| "detectionMethod": "MetricsStaticCondition", | ||
| "triggerType": "Critical", | ||
| "timeRange": "-5m", | ||
| "threshold": 80, | ||
| "thresholdType": "GreaterThanOrEqual", | ||
| "occurrenceType": "Always" | ||
| }, | ||
| { | ||
| "detectionMethod": "MetricsStaticCondition", | ||
| "triggerType": "ResolvedCritical", | ||
| "timeRange": "-5m", | ||
| "threshold": 80, | ||
| "thresholdType": "LessThan", | ||
| "occurrenceType": "Always" | ||
| } | ||
| ], | ||
| "notifications": [], | ||
| "isDisabled": true, | ||
| "groupNotifications": true, | ||
| "playbook": "" | ||
| }, | ||
| { | ||
| "name": "Couchbase - Node Down", | ||
| "description": "This alert fires when a node in the Couchbase cluster is down.", | ||
| "type": "MonitorsLibraryMonitorExport", | ||
| "monitorType": "Logs", | ||
| "evaluationDelay": "0m", | ||
| "alertName": null, | ||
| "runAs": null, | ||
| "notificationGroupFields": [], | ||
| "queries": [ | ||
| { | ||
| "rowId": "A", | ||
| "query": "db_cluster=* db_system=\"couchbase\" \"error\" \"nodedown\"\n| json \"log\" as _rawlog nodrop\n| if(isEmpty(_rawlog),_raw,_rawlog) as _raw\n| if (isEmpty(pod),_sourceHost,pod) as host\n|replace (_raw,/\\s+/,\" \") as _raw\n|parse regex \"nodedown,\\s'\\S+@(?<node>\\S+)\\'\"\n| parse regex \"ns_server:error,(?<time>\\S+),\"\n|_raw as msg \n| fields db_cluster,node" | ||
| } | ||
| ], | ||
| "triggers": [ | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "Critical", | ||
| "timeRange": "-5m", | ||
| "threshold": 0, | ||
| "thresholdType": "GreaterThan", | ||
| "field": "" | ||
| }, | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "ResolvedCritical", | ||
| "timeRange": "-5m", | ||
| "threshold": 0, | ||
| "thresholdType": "LessThanOrEqual", | ||
| "field": "" | ||
| } | ||
| ], | ||
| "notifications": [], | ||
| "isDisabled": true, | ||
| "groupNotifications": true, | ||
| "playbook": "" | ||
| }, | ||
| { | ||
| "name": "Couchbase - Node Not Respond", | ||
| "description": "This alert fires when a node in the Couchbase cluster does not respond too many times.", | ||
| "type": "MonitorsLibraryMonitorExport", | ||
| "monitorType": "Logs", | ||
| "evaluationDelay": "0m", | ||
| "alertName": null, | ||
| "runAs": null, | ||
| "notificationGroupFields": [], | ||
| "queries": [ | ||
| { | ||
| "rowId": "A", | ||
| "query": "db_cluster=* db_system=\"couchbase\" \"error\" \"Some nodes didn't respond\" \n| json \"log\" as _rawlog nodrop\n| if(isEmpty(_rawlog),_raw,_rawlog) as _raw\n|replace (_raw,/\\s+/,\" \") as _raw\n| parse regex \"stats:error,(?<time>\\S+),\"\n| parse regex \"Some nodes didn't respond: \\[(?<temp_nodes>.+)\\]\" | parse regex field=temp_nodes \"\\'(?<node_temp>[^,]+)\\'\" multi | parse regex field=node_temp \"@(?<node>.+)\"| _raw as msg \n| count as count by db_cluster,node" | ||
| } | ||
| ], | ||
| "triggers": [ | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "Critical", | ||
| "timeRange": "-5m", | ||
| "threshold": 10, | ||
| "thresholdType": "GreaterThanOrEqual", | ||
| "field": "count" | ||
| }, | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "ResolvedCritical", | ||
| "timeRange": "-5m", | ||
| "threshold": 10, | ||
| "thresholdType": "LessThan", | ||
| "field": "count" | ||
| } | ||
| ], | ||
| "notifications": [], | ||
| "isDisabled": true, | ||
| "groupNotifications": true, | ||
| "playbook": "" | ||
| }, | ||
| { | ||
| "name": "Couchbase - Bucket Not Ready", | ||
| "description": "This alert fires when a bucket in the Couchbase cluster is not ready.", | ||
| "type": "MonitorsLibraryMonitorExport", | ||
| "monitorType": "Logs", | ||
| "evaluationDelay": "0m", | ||
| "alertName": null, | ||
| "runAs": null, | ||
| "notificationGroupFields": [], | ||
| "queries": [ | ||
| { | ||
| "rowId": "A", | ||
| "query": "db_cluster=* db_system=\"couchbase\" \"buckets became not ready on node\" \"error\"\n| json \"log\" as _rawlog nodrop\n| if(isEmpty(_rawlog),_raw,_rawlog) as _raw\n| if (isEmpty(pod),_sourceHost,pod) as host\n|replace (_raw,/\\s+/,\" \") as _raw\n| parse regex \"\\'\\S+@(?<node>\\S+)\\'\\:\\s+\\[(?<buckets>.+)\\],\" | parse regex field=buckets \"\\\"(?<bucket>[^,]+)\\\"\" multi\n| parse regex \"ns_server:error,(?<time>\\S+),\" | _raw as msg\n| fields db_cluster,bucket,node" | ||
| } | ||
| ], | ||
| "triggers": [ | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "Critical", | ||
| "timeRange": "-5m", | ||
| "threshold": 0, | ||
| "thresholdType": "GreaterThan", | ||
| "field": "" | ||
| }, | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "ResolvedCritical", | ||
| "timeRange": "-5m", | ||
| "threshold": 0, | ||
| "thresholdType": "LessThanOrEqual", | ||
| "field": "" | ||
| } | ||
| ], | ||
| "notifications": [], | ||
| "isDisabled": true, | ||
| "groupNotifications": true, | ||
| "playbook": "" | ||
| }, | ||
| { | ||
| "name": "Couchbase - Too Many Error Queries on Buckets", | ||
| "description": "This alert fires when there are too many error queries on a bucket in a Couchbase cluster.", | ||
| "type": "MonitorsLibraryMonitorExport", | ||
| "monitorType": "Logs", | ||
| "evaluationDelay": "0m", | ||
| "alertName": null, | ||
| "runAs": null, | ||
| "notificationGroupFields": [], | ||
| "queries": [ | ||
| { | ||
| "rowId": "A", | ||
| "query": "db_cluster=* db_system=\"couchbase\" (\"ERROR\" or \"Error\")\n| json \"log\" as _rawlog nodrop\n| if(isEmpty(_rawlog),_raw,_rawlog) as _raw\n| if (isEmpty(pod),_sourceHost,pod) as host\n| parse regex \"_time=(?<time>\\S+)\"\n| parse regex \"_msg=(?<msg>.+)\"\n| parse regex field=msg \"Keyspace\\s\\w+:(?<bucket>.+)\\.\"\n| parse regex field=msg \"Failed to perform (?<method>\\w+)\"\n|count by db_cluster,host,bucket\n| fields db_cluster,host,bucket\n" | ||
| } | ||
| ], | ||
| "triggers": [ | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "Critical", | ||
| "timeRange": "-5m", | ||
| "threshold": 1000, | ||
| "thresholdType": "GreaterThanOrEqual", | ||
| "field": "" | ||
| }, | ||
| { | ||
| "detectionMethod": "LogsStaticCondition", | ||
| "triggerType": "ResolvedCritical", | ||
| "timeRange": "-5m", | ||
| "threshold": 1000, | ||
| "thresholdType": "LessThan", | ||
| "field": "" | ||
| } | ||
| ], | ||
| "notifications": [], | ||
| "isDisabled": true, | ||
| "groupNotifications": true, | ||
| "playbook": "" | ||
| } | ||
| ] | ||
| } |
Oops, something went wrong.