ApsaraDB for Redis monitors more than 10 groups of metrics in real time to help you understand the running status of ApsaraDB for Redis instances. This topic describes the purpose of each metric.

Monitoring frequency and monitoring cycle

The frequency that the performance monitoring system collects monitoring data is the monitoring frequency. If the system collects data every 5 seconds, the monitoring frequency is 5 seconds/time, and a monitoring cycle (or a collection cycle) is 5 seconds. For more information about how to modify the monitoring frequency, see Modify the monitoring frequency.

Note The default monitoring frequency is 60 seconds/time.

Monitoring group

You can switch between different monitoring groups. For more information, see Customize metrics. The following table describes the monitoring groups.

Monitoring group Description
Basic monitoring group The basic instance monitoring information, such as the queries per second (QPS), bandwidth, and memory usage.
Key monitoring group The monitoring information on the use of key-value related commands, such as the number of times DEL and EXITS are called.
String monitoring group The monitoring information on the use of string-related commands, such as the number of times APPEND and MGET are called.
Hash monitoring group The monitoring information on the use of hash-related commands, such as the number of times HGET and HDEL are called.
List monitoring group The monitoring information on the use of list-related commands, such as the number of times BLPOP and BRPOP are called.
Set monitoring group The monitoring information on the use of set-related commands, such as the number of times SADD and SCARD are called.
Zset monitoring group The monitoring information on the use of zset-related commands, such as the number of times ZADD and ZCARD are called.
HyperLog monitoring group The monitoring information on the use of HyperLogLog-related commands, such as the number of times PFADD and PFCOUNT are called.
Pub/Sub monitoring group The monitoring information on the use of publication and subscription-related commands, such as the number of times PUBLISH and SUBSCRIBE are called.
Transaction monitoring group The monitoring information on the use of transaction-related commands, such as the number of times WATCH, MULTI, and EXEC are called.
Lua script monitoring group The monitoring information on the use of Lua script-related commands, such as the number of times EVAL and SCRIPT are called.

Metrics in the basic monitoring group

Metric Unit Description Statistical method
CpuUsage % The CPU usage. Check the CPU usage when collecting monitoring data.
UsedMemory Bytes The size of the used memory. Check the memory usage when collecting monitoring data.
TotalQps Counts/s The number of requests received by the instance per second. Divide the number of requests in a monitoring cycle by the number of seconds in the monitoring cycle.
ConnCount Counts The number of connections. Check the number of connections when collecting monitoring data.
InFlow kbit/s The amount of data received by the instance per second. Divide the amount of data received in a monitoring cycle by the number of seconds in the monitoring cycle.
OutFlow kbit/s The amount of data sent by the instance per second. Divide the amount of data sent in a monitoring cycle by the number of seconds in the monitoring cycle.
FailedCount Counts/s The average number of abnormal requests per second. Divide the total number of abnormal requests in a monitoring cycle by the number of seconds in the monitoring cycle.
AvgRt µs The average response time of all requests.
Note For more information, see Response time metrics.
Divide the processing time of all requests in a monitoring cycle by the number of requests in the monitoring cycle.
MaxRt µs The maximum response time of requests.
Note For more information, see Response time metrics.
Check the maximum time consumed for processing a single request in a monitoring cycle.
Keys Counts The total number of keys. Check the number of keys when collecting monitoring data.
Expires Counts The total number of keys for which an expiration time is set. Check the cumulative value when collecting monitoring data. If the instance is restarted, the cumulative value is calculated again.
ExpiredKeys Counts The total number of expired keys. Check the cumulative value when collecting monitoring data. If the instance is restarted, the cumulative value is calculated again.
EvictedKeys Counts The total number of keys that are evicted because the memory is fully occupied. Check the cumulative value when collecting monitoring data. If the instance is restarted, the cumulative value is calculated again.
request Bytes The total amount of request data received by ApsaraDB for Redis nodes in a monitoring cycle. See the description of this metric.
response Bytes The total amount of response data sent by ApsaraDB for Redis nodes in a monitoring cycle. See the description of this metric.
request_max Bytes The maximum amount of data of a single request in a monitoring cycle. See the description of this metric.
response_max Bytes The maximum amount of data of a single response in a monitoring cycle. See the description of this metric.
traffic_control_input Counts The number of times that downlink throttling is triggered. Check the cumulative value in the monitoring cycle.
traffic_control_output Counts The number of times that uplink throttling is triggered. Check the cumulative value in the monitoring cycle.
traffic_control_input_status Counts Indicates whether downlink throttling has been triggered in a monitoring cycle. A value of 0 indicates that throttling has not been triggered, and a value of 1 indicates that throttling has been triggered. See the description of this metric.
traffic_control_output_status Counts Indicates whether uplink throttling has been triggered in a monitoring cycle. A value of 0 indicates that throttling has not been triggered, and a value of 1 indicates that throttling has been triggered. See the description of this metric.
hit_rate % The request hit rate, that is, the probability that data exists in an ApsaraDB for Redis instance for a data access request. Calculate the percentage of the hit requests to the total number of requests in a monitoring cycle.
hit Counts The number of hit requests. Check the number of hit requests in a monitoring cycle.
miss Counts The number of missed requests. Check the number of missed requests in a monitoring cycle.
evicted_keys_per_sec Counts/s The number of keys that are evicted per second. Divide the total number of keys evicted in a monitoring cycle by the number of seconds in the monitoring cycle.

Metrics in other monitoring groups

Besides the basic monitoring group, other monitoring groups contain metrics related to data of specific types or specific features. The metrics in these monitoring groups are divided into the following types:

  • Metrics that reflect the number of times commands are called. For example, the del, dump, and exists metrics in the key monitoring group are used to monitor the number of times the DEL, DUMP, and EXISTS commands are called.

  • Response time metrics of commands. For example, the metrics such as del_avg_rt, dump_avg_rt, and exists_avg_rt that are ended with avg_rt in the key monitoring group are used to monitor the average response time of the DEL, DUMP, and EXISTS commands in a monitoring cycle.

To understand how to use a metric, you only need to identify the metric category based on the metric name.

Response time metrics

All monitoring groups include response time metrics. Such metrics are ended with Rt or rt, for example, the AvgRt and MaxRt metrics in the basic monitoring group or the del_avg_rt and exists_avg_rt metrics in the key monitoring group.

The AvgRt and MaxRt metrics in the basic monitoring group are the most commonly used response time metrics. These two metrics have different meanings for proxy nodes and data nodes.

  • In an instance of the cluster edition or read/write splitting edition, the AvgRt metric of a proxy node reflects the average time consumed by the proxy node to process a command. A proxy node processes a command as follows:
    1. The proxy node receives a command and forwards it to a data node.
    2. The data node processes the command and replies to the proxy node.
    3. The proxy node returns the processing result.
    The AvgRt metric of the proxy node includes the time for the data node to process the command, the waiting time, and the time consumed for network communication between the proxy node and the data node.
  • In the data nodes of an instance of the cluster edition or read/write splitting edition or in an instance of the standard edition, the AvgRt metric reflects the average time consumed by a data node to process a command. This metric records the period from the time when the data node receives the command to the time when the data node returns the processing result. This metric does not include the time consumed by the proxy node to process the command and the time consumed for network communication.
  • The MaxRt metric indicates the maximum response time of requests. It is collected in the same way as the AvgRt metric in ApsaraDB for Redis instances of all editions.