Mackerel supports obtaining and monitoring Amazon Redshift metrics in AWS Integration. When integrating with AWS Integration, billable targets are determined using the conversion 1 Cluster = 1 Micro Host. In addition to this, depending on the number of metrics retrieved, you may be charged for exceeding the maximum number of metrics per micro host.
Please refer to the following page for AWS Integration configuration methods and a list of supported AWS services.
AWS Integration
Obtaining metrics
The metrics obtainable with AWS Integration’s Redshift support are as follows. For Metric explanations, refer to the AWS help page.
The maximum number of metrics obtainable is 24 + 2 × (number of queues) + 1 × (number of service classes) + 10 × (number of nodes).
Metrics per Cluster
The WLM_ID of the metric name contains the workload management (WLM) queue ID. (Example: 1, Default etc.)
The SERVICE_CLASS of the metric name contains the workload management (WLM) service class ID. (Example: 6, 7 etc.)
| Graph name | Metric | Metric name in Mackerel | Unit | Statistics |
|---|---|---|---|---|
| CPU | CPUUtilization | redshift.cpu.used | percentage | Average |
| Database Connections | DatabaseConnections | redshift.database_connections.used | float | Average |
| Cluster Status | HealthStatus MaintenanceMode |
redshift.cluster_status.health redshift.cluster_status.maintenance |
float | Average |
| Network Throughput | NetworkReceiveThroughput NetworkTransmitThroughput |
redshift.network_throughput.receive redshift.network_throughput.transmit |
bytes/sec | Average |
| Disk Space | PercentageDiskSpaceUsed | redshift.disk.used | percentage | Average |
| Total Table Count | TotalTableCount | redshift.total_table_count.count | float | Average |
| Query Runtime Breakdown | QueryRuntimeBreakdown | redshift.query_runtime_breakdown.planning redshift.query_runtime_breakdown.waiting redshift.query_runtime_breakdown.executing_read redshift.query_runtime_breakdown.executing_insert redshift.query_runtime_breakdown.executing_delete redshift.query_runtime_breakdown.executing_update redshift.query_runtime_breakdown.executing_ctas redshift.query_runtime_breakdown.executing_unload redshift.query_runtime_breakdown.executing_copy redshift.query_runtime_breakdown.commit |
float | Average |
| Query Throughput | QueriesCompletedPerSecond | redshift.query_throughput.short redshift.query_throughput.medium redshift.query_throughput.long |
float | Average |
| Query Duration | QueryDuration | redshift.query_duration.short redshift.query_duration.medium redshift.query_duration.long |
float | Average |
| WLM Query Throughput | WLMQueriesCompletedPerSecond | redshift.wlm_query_throughput.WLM_ID | float | Average |
| WLM Query Duration | WLMQueryDuration | redshift.wlm_query_duration.WLM_ID | float | Average |
| WLM Queue Length | WLMQueueLength | redshift.wlm_queue_length.SERVICE_CLASS | float | Average |
Metrics per Node
Since there can be multiple Nodes per cluster in Redshift, each metric is grouped as follows. The metric name’s NODE_ROLE will contain the Role of the Node. (Example: leader, compute_0 etc.)
| Graph name | Metric | Metric name in Mackerel | Unit | Statistics |
|---|---|---|---|---|
| CPU per Node | CPUUtilization | redshift.cpu_per_node.NODE_ROLE.used | percentage | Average |
| Network Throughput per Node | NetworkReceiveThroughput NetworkTransmitThroughput |
redshift.network_throughput_per_node.NODE_ROLE.receive redshift.network_throughput_per_node.NODE_ROLE.transmit |
bytes/sec | Average |
| Disk Space per Node | PercentageDiskSpaceUsed | redshift.disk_per_node.NODE_ROLE.used | percentage | Average |
| Disk IOPS | ReadIOPS WriteIOPS |
redshift.diskiops.NODE_ROLE.read redshift.diskiops.NODE_ROLE.write |
iops | Average |
| Disk Latency | ReadLatency WriteLatency |
redshift.latency.NODE_ROLE.read redshift.latency.NODE_ROLE.write |
float | Average |
| Disk Throughput | ReadThroughput WriteThroughput |
redshift.throughput.NODE_ROLE.read redshift.throughput.NODE_ROLE.write |
bytes/sec | Average |
Notes
Among the graphs/metrics obtainable with AWS integration mentioned above, the metric retrieval interval differs for metrics included in the following graph.
- 5 minute interval
- Query Runtime Breakdown
- Query Duration
- Query Throughput
- WLM Query Duration
- WLM Query Throughput
