Metrics for Valkey and Redis OSS - Amazon ElastiCache
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Metrics for Valkey and Redis OSS

The Amazon ElastiCache namespace includes the following Valkey and Redis OSS metrics. These metrics are the same when using the Valkey engine.

With the exception of ReplicationLag, EngineCPUUtilization, SuccessfulWriteRequestLatency, and SuccessfulReadRequestLatency, these metrics are derived from the info command. Each metric is calculated at the cache node level.

For complete documentation of the info command, see http://valkey.io/commands/info.

See Also

Metric Description Unit
ActiveDefragHits The number of value reallocations per minute performed by the active defragmentation process. This is derived from active_defrag_hits statistic at INFO. Number
AuthenticationFailures The total number of failed attempts to authenticate to Valkey or Redis OSS using the AUTH command. You can find more information about individual authentication failures using the ACL LOG command. We suggest setting an alarm on this to detect unauthorized access attempts. Count
BytesUsedForCache The total number of bytes allocated by Valkey or Redis OSS for all purposes, including the dataset, buffers, and so on. Bytes
Dimension: Tier=Memory for Valkey or Redis OSS clusters using Data tiering in ElastiCache: The total number of bytes used for cache by memory. This is the value of used_memory statistic at INFO. Bytes
Dimension: Tier=SSD for Valkey or Redis OSS clusters using Data tiering in ElastiCache: The total number of bytes used for cache by SSD. Bytes
BytesReadFromDisk The total number of bytes read from disk per minute. Supported only for clusters using Data tiering in ElastiCache. Bytes
BytesWrittenToDisk The total number of bytes written to disk per minute. Supported only for clusters using Data tiering in ElastiCache. Bytes
CacheHits The number of successful read-only key lookups in the main dictionary. This is derived from keyspace_hits statistic at INFO. Count
CacheMisses The number of unsuccessful read-only key lookups in the main dictionary. This is derived from keyspace_misses statistic at INFO. Count
CommandAuthorizationFailures The total number of failed attempts by users to run commands they don’t have permission to call. You can find more information about individual authentication failures using the ACL LOG command. We suggest setting an alarm on this to detect unauthorized access attempts. Count
CacheHitRate Indicates the usage efficiency of the Valkey or Redis OSS instance. If the cache ratio is lower than about 0.8, it means that a significant amount of keys are evicted, expired, or don't exist. This is calculated using cache_hits and cache_misses statistics in the following way: cache_hits /(cache_hits + cache_misses). Percent
ChannelAuthorizationFailures The total number of failed attempts by users to access channels they do not have permission to access. You can find more information about individual authentication failures using the ACL LOG command. We suggest setting an alarm on this metric to detect unauthorized access attempts. Count
CurrConnections The number of client connections, excluding connections from read replicas. ElastiCache uses 4 to 6 of the connections to monitor the cluster in each case. This is derived from the connected_clients statistic at INFO. Count
CurrItems The number of items in the cache. This is derived from the keyspace statistic, summing all of the keys in the entire keyspace. Count
Dimension: Tier=Memory for clusters using Data tiering in ElastiCache. The number of items in memory. Count
Dimension: Tier=SSD (solid state drives) for clusters using Data tiering in ElastiCache. The number of items in SSD. Count
CurrVolatileItems Total number of keys in all databases that have a ttl set. This is derived from the expires statistic, summing all of the keys with a ttl set in the entire keyspace. Count
DatabaseCapacityUsagePercentage

Percentage of the total data capacity for the cluster that is in use.

On Data Tiered instances, the metric is calculated as (used_memory - mem_not_counted_for_evict + SSD used) / (maxmemory + SSD total capacity), where used_memory and maxmemory are taken from INFO.

In all other cases, the metric is calculated using used_memory/maxmemory.

Percent
DatabaseCapacityUsageCountedForEvictPercentage

Percentage of the total data capacity for the cluster that is in use, excluding the memory used for overhead and COB. This metric is calculated as:

used_memory - mem_not_counted_for_evict/maxmemory

On Data Tiered instances, the metric is calculated as:

(used_memory + SSD used) / (maxmemory + SSD total capacity)

where used_memory and maxmemory are taken from INFO

Percent
DatabaseMemoryUsagePercentage Percentage of the memory for the cluster that is in use. This is calculated using used_memory/maxmemory from INFO. Percent
DatabaseMemoryUsageCountedForEvictPercentage Percentage of the memory for the cluster that is in use, excluding memory used for overhead and COB. This is calculated using used_memory-mem_not_counted_for_evict/maxmemory from INFO. Percent
DB0AverageTTL Exposes avg_ttl of DBO from the keyspace statistic of INFO command. Replicas don't expire keys, instead they wait for primary nodes to expire keys. When a primary node expires a key (or evicts it because of LRU), it synthesizes a DEL command, which is transmitted to all the replicas. Therefore, DB0AverageTTL is 0 for replica nodes, due the fact that they don't expire keys, and thus don't track TTL. Milliseconds
EngineCPUUtilization

Provides CPU utilization of the Valkey or Redis OSS engine thread. Because Valkey and Redis OSS are single-threaded, you can use this metric to analyze the load of the process itself. The EngineCPUUtilization metric provides a more precise visibility of the process. You can use it in conjunction with the CPUUtilization metric. CPUUtilization exposes CPU utilization for the server instance as a whole, including other operating system and management processes. For larger node types with four vCPUs or more, use the EngineCPUUtilization metric to monitor and set thresholds for scaling.

Note

On an ElastiCache host, background processes monitor the host to provide a managed database experience. These background processes can take up a significant portion of the CPU workload. This is not significant on larger hosts with more than two vCPUs. But it can affect smaller hosts with 2vCPUs or fewer. If you only monitor the EngineCPUUtilization metric, you will be unaware of situations where the host is overloaded with both high CPU usage from Valkey or Redis OSS and high CPU usage from the background monitoring processes. Therefore, we recommend monitoring the CPUUtilization metric for hosts with two vCPUs or less.

Percent
Evictions The number of keys that have been evicted due to the maxmemory limit. This is derived from the evicted_keys statistic at INFO. Count
GlobalDatastoreReplicationLag This is the lag between the secondary Region's primary node and the primary Region's primary node. For cluster mode enabled Valkey or Redis OSS, the lag indicates the maximum delay among the shards. Seconds
IamAuthenticationExpirations The total number of expired IAM-authenticated Valkey or Redis OSS connections. You can find more information about Authenticating with IAM in the user guide. Count
IamAuthenticationThrottling The total number of throttled IAM-authenticated Valkey or Redis OSS AUTH or HELLO requests. You can find more information about Authenticating with IAM in the user guide. Count
IsMaster Indicates whether the node is the primary node of current shard/cluster. The metric can be either 0 (not primary) or 1 (primary). Count
KeyAuthorizationFailures The total number of failed attempts by users to access keys they don’t have permission to access. You can find more information about individual authentication failures using the ACL LOG command. We suggest setting an alarm on this to detect unauthorized access attempts. Count
KeysTracked The number of keys being tracked by Valkey or Redis OSS key tracking as a percentage of tracking-table-max-keys. Key tracking is used to aid client-side caching and notifies clients when keys are modified. Count
MemoryFragmentationRatio Indicates the efficiency in the allocation of memory of the Valkey or Redis OSS engine. Certain thresholds signify different behaviors. The recommended value is to have fragmentation above 1.0. This is calculated from the mem_fragmentation_ratio statistic of INFO. Number
NewConnections The total number of connections that have been accepted by the server during this period. This is derived from the total_connections_received statistic at INFO.
Note

If you are using ElastiCache (Redis OSS) version 5 or lower, between two and four of the connections reported by this metric are used by ElastiCache to monitor the cluster. However, when using ElastiCache (Redis OSS) version 6 or above, the connections used by ElastiCache to monitor the cluster are not included in this metric.

Count
NumItemsReadFromDisk The total number of items retrieved from disk per minute. Supported only for clusters using Data tiering in ElastiCache. Count
NumItemsWrittenToDisk The total number of items written to disk per minute. Supported only for clusters using Data tiering in ElastiCache. Count
MasterLinkHealthStatus This status has two values: 0 or 1. The value 0 indicates that data in the ElastiCache primary node is not in sync with Valkey or Redis OSS on EC2. The value of 1 indicates that the data is in sync. To complete the migration, use the CompleteMigration API operation. Boolean
Reclaimed The total number of key expiration events. This is derived from the expired_keys statistic at INFO. Count
ReplicationBytes For nodes in a replicated configuration, ReplicationBytes reports the number of bytes that the primary is sending to all of its replicas. This metric is representative of the write load on the replication group. This is derived from the master_repl_offset statistic at INFO. Bytes
ReplicationLag This metric is only applicable for a node running as a read replica. It represents how far behind, in seconds, the replica is in applying changes from the primary node. For Valkey 7.2 and onwards, and Redis OSS engine version 5.0.6 onwards, the lag can be measured in milliseconds. Seconds
SaveInProgress This binary metric returns 1 whenever a background save (forked or forkless) is in progress, and 0 otherwise. A background save process is typically used during snapshots and syncs. These operations can cause degraded performance. Using the SaveInProgress metric, you can diagnose whether degraded performance was caused by a background save process. This is derived from the rdb_bgsave_in_progress statistic at INFO. Boolean
TrafficManagementActive Indicates whether ElastiCache (Redis OSS) is actively managing traffic by adjusting traffic allocated to incoming commands, monitoring or replication. Traffic is managed when more commands are sent to the node than can be processed by Valkey or Redis OSS and is used to maintain the stability and optimal operation of the engine. Any data points of 1 may indicate that the node is underscaled for the workload being provided.
Note

If this metric remains active, evaluate the cluster to decide if scaling up or scaling out is necessary. Related metrics include NetworkBandwidthOutAllowanceExceeded and EngineCPUUtilization.

Boolean
SuccessfulWriteRequestLatency

Latency of successful write requests.

Valid statistics: Average, Sum, Min, Max, Sample Count, any percentile between p0 and p100. The sample count includes only the commands that were successfully executed.

Microseconds
SuccessfulReadRequestLatency

Latency of successful read requests.

Valid statistics: Average, Sum, Min, Max, Sample Count, any percentile between p0 and p100. The sample count includes only the commands that were successfully executed.

Microseconds
ErrorCount

The total number of failed commands during the specified time period.

Valid statistics: Average, Sum, Min, Max

Count
EngineCPUUtilization availability

Amazon Regions listed following are available on all supported node types.

Region Region name
us-east-2 US East (Ohio)
us-east-1 US East (N. Virginia)
us-west-1 US West (N. California)
us-west-2 US West (Oregon)
ap-northeast-1 Asia Pacific (Tokyo)
ap-northeast-2 Asia Pacific (Seoul)
ap-northeast-3 Asia Pacific (Osaka)
ap-east-1 Asia Pacific (Hong Kong)
ap-south-1 Asia Pacific (Mumbai)
ap-southeast-1 Asia Pacific (Singapore)
ap-southeast-2 Asia Pacific (Sydney)
ap-southeast-3 Asia Pacific (Jakarta)
ca-central-1 Canada (Central)
cn-north-1 China (Beijing)
cn-northwest-2 China (Ningxia)
me-south-1 Middle East (Bahrain)
eu-central-1 Europe (Frankfurt)
eu-west-1 Europe (Ireland)
eu-west-2 Europe (London)
eu-west-3 EU (Paris)
eu-south-1 Europe (Milan)
af-south-1 Africa (Cape Town)
eu-north-1 Europe (Stockholm)
sa-east-1 South America (São Paulo)

The following are aggregations of certain kinds of commands, derived from info commandstats. The commandstats section provides statistics based on the command type, including the number of calls, the total CPU time consumed by these commands, and the average CPU consumed per command execution. For each command type, the following line is added: cmdstat_XXX: calls=XXX,usec=XXX,usec_per_call=XXX.

The latency metrics listed following are calculated using commandstats statistic from INFO. They are calculated in the following way: delta(usec)/delta(calls). delta is calculated as the diff within one minute. Latency is defined as CPU time taken by ElastiCache to process the command. Note that for clusters using data tiering, the time taken to fetch items from SSD is not included in these measurements.

For a full list of available commands, see commands in the Valkey documentation.

Metric Description Unit
ClusterBasedCmds The total number of commands that are cluster-based. This is derived from the commandstats statistic by summing all of the commands that act upon a cluster (cluster slot, cluster info, and so on). Count
ClusterBasedCmdsLatency Latency of cluster-based commands. Microseconds
EvalBasedCmds The total number of commands for eval-based commands. This is derived from the commandstats statistic by summing eval, evalsha. Count
EvalBasedCmdsLatency Latency of eval-based commands. Microseconds
GeoSpatialBasedCmds The total number of commands for geospatial-based commands. This is derived from the commandstats statistic. It's derived by summing all of the geo type of commands: geoadd, geodist, geohash, geopos, georadius, and georadiusbymember. Count
GeoSpatialBasedCmdsLatency Latency of geospatial-based commands. Microseconds
GetTypeCmds The total number of read-only type commands. This is derived from the commandstats statistic by summing all of the read-only type commands (get, hget, scard, lrange, and so on.) Count
GetTypeCmdsLatency Latency of read commands. Microseconds
HashBasedCmds The total number of commands that are hash-based. This is derived from the commandstats statistic by summing all of the commands that act upon one or more hashes (hget, hkeys, hvals, hdel, and so on). Count
HashBasedCmdsLatency Latency of hash-based commands. Microseconds
HyperLogLogBasedCmds The total number of HyperLogLog-based commands. This is derived from the commandstats statistic by summing all of the pf type of commands (pfadd, pfcount, pfmerge, and so on.). Count
HyperLogLogBasedCmdsLatency Latency of HyperLogLog-based commands. Microseconds
JsonBasedCmds The total number of JSON commands, including both read and write commands. This is derived from the commandstats statistic by summing all JSON commands that act upon JSON keys. Count
JsonBasedCmdsLatency Latency of all JSON commands, including both read and write commands. Microseconds
JsonBasedGetCmds The total number of JSON read-only commands. This is derived from the commandstats statistic by summing all JSON read commands that act upon JSON keys. Count
JsonBasedGetCmdsLatency Latency of JSON read-only commands. Microseconds
JsonBasedSetCmds The total number of JSON write commands. This is derived from the commandstats statistic by summing all JSON write commands that act upon JSON keys. Count
JsonBasedSetCmdsLatency Latency of JSON write commands. Microseconds
KeyBasedCmds The total number of commands that are key-based. This is derived from the commandstats statistic by summing all of the commands that act upon one or more keys across multiple data structures (del, expire, rename, and so on.). Count
KeyBasedCmdsLatency Latency of key-based commands. Microseconds
ListBasedCmds The total number of commands that are list-based. This is derived from the commandstats statistic by summing all of the commands that act upon one or more lists (lindex, lrange, lpush, ltrim, and so on). Count
ListBasedCmdsLatency Latency of list-based commands. Microseconds
NonKeyTypeCmds The total number of commands that are not key-based. This is derived from the commandstats statistic by summing all of the commands that do not act upon a key, for example, acl, dbsize or info. Count
NonKeyTypeCmdsLatency Latency of non-key-based commands. Microseconds
PubSubBasedCmds The total number of commands for pub/sub functionality. This is derived from the commandstatsstatistics by summing all of the commands used for pub/sub functionality: psubscribe, publish, pubsub, punsubscribe, ssubscribe, sunsubscribe, spublish, subscribe, and unsubscribe. Count
PubSubBasedCmdsLatency Latency of pub/sub-based commands. Microseconds
SetBasedCmds The total number of commands that are set-based. This is derived from the commandstats statistic by summing all of the commands that act upon one or more sets (scard, sdiff, sadd, sunion, and so on). Count
SetBasedCmdsLatency Latency of set-based commands. Microseconds
SetTypeCmds The total number of write types of commands. This is derived from the commandstats statistic by summing all of the mutative types of commands that operate on data (set, hset, sadd, lpop, and so on.) Count
SetTypeCmdsLatency Latency of write commands. Microseconds
SortedSetBasedCmds The total number of commands that are sorted set-based. This is derived from the commandstats statistic by summing all of the commands that act upon one or more sorted sets (zcount, zrange, zrank, zadd, and so on). Count
SortedSetBasedCmdsLatency Latency of sorted-based commands. Microseconds
StringBasedCmds The total number of commands that are string-based. This is derived from the commandstats statistic by summing all of the commands that act upon one or more strings (strlen, setex, setrange, and so on). Count
StringBasedCmdsLatency Latency of string-based commands. Microseconds
StreamBasedCmds The total number of commands that are stream-based. This is derived from the commandstats statistic by summing all of the commands that act upon one or more streams data types (xrange, xlen, xadd, xdel, and so on). Count
StreamBasedCmdsLatency Latency of stream-based commands. Microseconds