GuanceDB Engine Deployment¶
Note
The CPU needs to support the AVX2 instruction set. Execute grep avx2 /proc/cpuinfo to test. If there is no output, AVX2 is not supported.
This document describes how to deploy the GuanceDB for Metrics component and the GuanceDB for Logs component. Both components are deployed using a single Chart.
Architecture Diagram¶
- The GuanceDB chart includes both the GuanceDB for Metrics component and the GuanceDB for Logs component. These two components share the guance-insert and guance-select services, providing a unified write and query interface.
- Deploying the GuanceDB for Metrics component will deploy the GuanceDB for Metrics Storage.
Core Components¶
1.Write Load Balancer (guance-insert)
- Responsible for receiving raw data.
- Distributes data to storage nodes using a consistent hashing algorithm.
- Supports high availability deployment.
2.Query Engine (guance-select)
- Supports DQL and PromQL query languages.
- Retrieves data from all storage nodes and performs calculations.
- Returns the final query results.
3.Storage Node (guance-storage)
- Uses a share nothing architecture.
- No communication between nodes, no central coordination.
- Supports horizontal scaling.
Prerequisites¶
- Helm version requirement: 3.7+
- Ensure storage performance meets requirements.
- It is recommended to reserve sufficient system resources.
Default Deployment Configuration Information¶
| Service Name | Address | Port |
|---|---|---|
| guance-insert | guance-insert.guancedb | 8480 |
| guance-select | guance-select.guancedb | 8481 |
Capacity Planning¶
1.Memory Reservation
- All components should reserve at least 50% of memory.
- Used to handle sudden write bursts and avoid OOM.
2.CPU Reservation
- All components should reserve at least 50% of CPU.
- Avoids slow writes and queries caused by sudden write bursts.
3.Storage Space Reservation
- guance-storage nodes should reserve 20% of storage space.
- Can be configured using the
-storage.minFreeDiskSpaceBytesparameter.
Reference Configuration¶
For 1 million time series (writing every 10s, real-time write QPS of 100k/s, query QPS of 100, data retention of 1 month):
| Component | CPU | Memory | Disk |
|---|---|---|---|
| guance-insert | 4c | 4G | - |
| guance-select | 8c | 8G | - |
| guance-storage | 16c | 96G | 500G |
Note: The resource scale can be adjusted linearly based on the number of time series during actual deployment. Each component supports 2 instances by default.
GuanceDB Engine Deployment¶
Note
The deployment of GuanceDB for Metrics and GuanceDB for Logs has a sequential relationship. You can deploy GuanceDB for Metrics first, and after Doris is deployed, you can configure Doris information and install GuanceDB for Logs. If you only want to deploy the GuanceDB for Logs component, you can set guance_storage.enabled=false.
Storage Class Settings¶
OpenEBS reference documentation: OpenEBS Deployment
If using a public cloud, please refer to the public cloud storage block component.
Deploy the following yaml configuration:
apiVersion: storage.k8s.io/v1
allowVolumeExpansion: true
kind: StorageClass
metadata:
annotations:
cas.openebs.io/config: |
- name: StorageType
value: "hostpath"
- name: BasePath
value: "/data"
name: guancedb-cs
provisioner: openebs.io/local
reclaimPolicy: Retain
volumeBindingMode: WaitForFirstConsumer
Ensure that the
/datadirectory has sufficient disk capacity.
Deploy the following yaml configuration:
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: guancedb-cs
provisioner: diskplugin.csi.alibabacloud.com
parameters:
type: cloud_auto
fstype: ext4
# diskTags: "a:b,b:c"
encrypted: "false"
performanceLevel: PL1 # When using with cloud box, set to PL0.
provisionedIops: "40000"
burstingEnabled: "false"
volumeBindingMode: WaitForFirstConsumer
reclaimPolicy: Retain
allowVolumeExpansion: true
Deploy the following yaml configuration:
---
allowVolumeExpansion: true
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: guancedb-cs
parameters:
csi.storage.k8s.io/csi-driver-name: disk.csi.everest.io
csi.storage.k8s.io/fstype: ext4
everest.io/disk-volume-type: SSD ## Modify disk type
everest.io/passthrough: "true"
provisioner: everest-csi-provisioner
reclaimPolicy: Delete
volumeBindingMode: WaitForFirstConsumer
Download Chart¶
Note
It is recommended to use Git to manage values.yaml, include the values.yaml file in Git management, ensure that each configuration change is recorded, and facilitate tracking and rollback.
Download the chart package:
--untar is for uncompressed, default is compressed.
Configure and Install GuanceDB for Metrics¶
1.Configure GuanceDB for Metrics
Enter the guancedb directory and modify values.yaml
# Modify the global storage class
global:
storageClass: 'guancedb-cs'
# Turn off the scopedb switch
scopedb:
enabled: false
guance_select:
...
# Number of service replicas
replicaCount: 3
image:
# If in an offline environment, modify the image address. If there is no such requirement, do not modify.
repository: pubrepo.guance.com/guancedb/guance-select
# Modify nodeSelector and tolerations to schedule guance-select containers. If there is no such requirement, do not modify.
nodeSelector: {}
tolerations:
[]
# - key: "key"
# operator: "Equal|Exists"
# value: "value"
# effect: "NoSchedule|PreferNoSchedule"
guance_insert:
...
# Number of service replicas
replicaCount: 3
image:
# If in an offline environment, modify the image address. If there is no such requirement, do not modify.
repository: pubrepo.guance.com/guancedb/guance-insert
# Modify nodeSelector and tolerations to schedule guance-insert containers. If there is no such requirement, do not modify.
nodeSelector: {}
tolerations:
[]
# - key: "key"
# operator: "Equal|Exists"
# value: "value"
# effect: "NoSchedule|PreferNoSchedule"
guance_storage:
...
# Number of service replicas
replicaCount: 2
image:
# If in an offline environment, modify the image address. If there is no such requirement, do not modify.
repository: pubrepo.guance.com/guancedb/guance-storage
# Modify nodeSelector and tolerations to schedule guance-storage containers. If there is no such requirement, do not modify.
nodeSelector: {}
tolerations:
[]
# - key: "key"
# operator: "Equal|Exists"
# value: "value"
# effect: "NoSchedule|PreferNoSchedule"
2.Install GuanceDB for Metrics
helm upgrade -i guancedb oci://pubrepo.guance.com/guancedb/guancedb \
-n guancedb --create-namespace -f values.yaml
3.Check Status {#helm-check}
> kubectl get pod -n guancedb -l app.kubernetes.io/instance=guancedb
NAME READY STATUS RESTARTS AGE
guance-insert-0 1/1 Running 0 107s
guance-select-db5dfb66b-qsncr 1/1 Running 0 84s
guance-storage-0 1/1 Running 0 44s
guance-storage-1 1/1 Running 0 28s
Configure and Install GuanceDB for Logs¶
Note
If GuanceDB for Metrics is already deployed, installing GuanceDB for Logs will not add new services, but only add service startup parameters.
1.Configure GuanceDB for Logs
Enter the guancedb directory and modify values.yaml
Note
values.yaml is the file for configuring GuanceDB for Metrics. If values.yaml is lost, you can execute helm get values -n guancedb guancedb > values.yaml to obtain it.
...
doris:
# Enable GuanceDB for Logs
enabled: true
# Add Doris fe nodes
feIPs: []
# ["xx.xx.xx.xx","xx.xx.xx.xx"]
# Doris root password
rootPassword: ''
# Doris regular user password
userReadPassword: ''
# memory size for cache data
cacheSizeLimit: 4Gi
guance_throttler:
# Enable
enabled: true
...
2.Install GuanceDB for Logs
helm upgrade -i guancedb oci://pubrepo.guance.com/guancedb/guancedb \
-n guancedb --create-namespace -f values.yaml
3.Check Status {#helm-check-logs}
> kubectl get pod -n guancedb -l app.kubernetes.io/instance=guancedb
NAME READY STATUS RESTARTS AGE
guance-insert-0 1/1 Running 0 107s
guance-select-db5dfb66b-qsncr 1/1 Running 0 84s
guance-storage-0 1/1 Running 0 44s
guance-storage-1 1/1 Running 0 28s
Uninstall¶
Other¶
Modify Component Parameters¶
1.Modify values.yaml
guance_storage:
...
# You can add storage parameters here
extraArgs:
retentionFilters.config: http://kodo.forethought-kodo:9527/v1/workspace_retention
envflag.enable: 'true'
loggerFormat: json
loggerMaxArgLen: '10000'
loggerTimezone: Asia/Shanghai
guance_insert:
...
# You can add insert parameters here
extraArgs:
envflag.enable: 'true'
loggerFormat: json
loggerMaxArgLen: '10000'
loggerTimezone: Asia/Shanghai
streamaggr.insertVerify: 'true'
guance_select:
# You can add select parameters here
...
extraArgs:
envflag.enable: 'true'
loggerFormat: json
loggerMaxArgLen: '10000'
loggerTimezone: Asia/Shanghai
http.disableResponseCompression: 'true'
search.denyPartialResponse: true
search.maxQueryDuration: 110s
slowQueryDuration: 1ms
streamaggr.selectVerify: 'true'
2.Update Deployment:
helm upgrade -i guancedb oci://pubrepo.guance.com/guancedb/guancedb \
-n guancedb --create-namespace -f values.yaml
Service Parameter Description¶
Guance Select Parameter Description¶
| Parameter Name | Type | Default Value | Description |
|---|---|---|---|
| -blockcache.missesBeforeCaching | int | 2 | The number of cache misses before a cache block is placed into the cache. A higher value may reduce the size of the indexdb/dataBlocks cache but increase CPU and disk read operations. |
| -cacheDataPath | string | "/tmp/guance-select" | The path to the cache file directory. If empty, no cache is saved. |
| -cacheExpireDuration | duration | 30m0s | The duration after which unaccessed cache items are removed from memory. A lower value may reduce memory usage but increase CPU usage. |
| -dedup.minScrapeInterval | duration | 1ms | Only the last sample within each discrete time interval is retained for each time series, preventing duplication. |
| -denyQueryTracing | bool | Whether to disable query tracing. | |
| -doris.coldDataTierOnly | bool | Use only cold storage. | |
| -doris.disableBFOnNewTables | bool | Do not use Bloom filters for new tables. | |
| -doris.disableUnicodeTokenizer | bool | Use a Chinese tokenizer instead of a Unicode tokenizer. | |
| -doris.enableBFOnOldTables | bool | Use Bloom filters for old tables. | |
| -doris.enableFileCacheTTL | bool | Use the data cache of the cloud database SelectDB version. | |
| -doris.enableWarmDataTier | bool | Use hot plus warm storage instead of hot plus cold storage. | |
| -doris.indexCacheTTL | duration | 30m0s | The TTL of the Doris index cache. |
| -doris.onlyQueryFE | bool | When enabled, do not query data directly from the backend (be). | |
| -doris.onlyWriteFE | bool | When enabled, do not write data directly to the backend (be). | |
| -doris.replicationFactor | int | 1 | The number of replicas for Doris tables. |
| -doris.storageResource | string | "s3_default" | The name of the storage policy resource used. |
| -doris.variantSchemaMode | string | Supported values: 'none', 'minimal', 'full'. | |
| -dorisCluster | string | "default_cluster" | The name of the Doris cluster. |
| -dorisFENode | array | The list of Doris FE node addresses, comma-separated. | |
| -dorisHTTPPort | int | 8030 | The HTTP port of Doris. |
| -dorisMySQLPort | int | 9030 | The MySQL port of Doris. |
| -dorisPassword | string | The password for Doris. | |
| -dorisUsername | string | "root" | The username for Doris. |
| -enableTCP6 | bool | Whether to enable IPv6 listening and dialing. By default, only IPv4 is used. | |
| -envflag.enable | bool | Whether to allow reading parameters from environment variables, with command-line parameters taking precedence. | |
| -envflag.prefix | string | The prefix for environment variables, which takes effect only when reading from environment variables is enabled. | |
| -filestream.disableFadvise | bool | Whether to disable the fadvise() system call, which can reduce CPU usage when reading large files but is generally not recommended to turn off. | |
| -flagsAuthKey | value | The authentication key for the /flags endpoint. Supported to be read from a file or URL. | |
| -fs.disableMmap | bool | Whether to use pread() instead of mmap() to read data files. By default, mmap() is used on 64-bit architectures and pread() is used on 32-bit architectures. | |
| -globalReplicationFactor | int | 1 | The number of replicas for each sample in the vmstorage group. Supports high availability. |
| -http.connTimeout | duration | 2m0s | The timeout for HTTP connections, after which the connection is closed. 0 means no timeout. |
| -http.disableResponseCompression | bool | Whether to disable HTTP response compression to save CPU. By default, compression is enabled to save bandwidth. | |
| -http.header.csp | string | Set the Content-Security-Policy header, recommended: "default-src 'self'". | |
| -http.header.frameOptions | string | Set the X-Frame-Options header. | |
| -http.header.hsts | string | Set the Strict-Transport-Security header, recommended: 'max-age=31536000; includeSubDomains'. | |
| -http.idleConnTimeout | duration | 1m0s | The timeout for idle HTTP connections. |
| -http.maxGracefulShutdownDuration | duration | 7s | The maximum duration for graceful shutdown of the HTTP server. |
| -http.pathPrefix | string | Add a prefix to all HTTP paths, suitable for proxy paths. | |
| -http.shutdownDelay | duration | The delay before the HTTP server shuts down, during which /health returns non-OK, making it easier for load balancers to switch requests. | |
| -httpAuth.password | value | The password for HTTP basic authentication, supported to be read from a file or URL. | |
| -httpAuth.username | string | The username for HTTP basic authentication. If empty, authentication is disabled. | |
| -httpListenAddr | array | The list of HTTP listening addresses, supports multiple values or comma-separated values. | |
| -httpListenAddr.useProxyProtocol | array | Whether to enable the proxy protocol for the specified listening address. When enabled, the /metrics endpoint cannot be provided. | |
| -insert.maxFutureTimeDrift | duration | 1h0m0s | Events with timestamps greater than the current time plus this value will be rejected. |
| -insert.maxPastTimeDrift | duration | Events with timestamps less than the current time minus this value will be rejected. | |
| -insertNode | array | The list of Guance-insert node addresses, comma-separated. | |
| -internStringCacheExpireDuration | duration | 6m0s | The TTL of the interned string cache. |
| -internStringDisableCache | bool | Whether to disable the interned string cache. Disabling it may reduce memory but increase CPU usage. | |
| -internStringMaxLen | int | 500 | The maximum length of interned strings. A lower value may reduce memory but increase CPU usage. |
| -loggerDisableTimestamps | bool | false | Whether to disable timestamp writing in logs. |
| -loggerErrorsPerSecondLimit | int | 0 | The limit on the number of error logs per second. Excess errors will be suppressed. 0 means no limit. |
| -loggerFormat | string | "default" | The log format, options: "default" or "json". |
| -loggerJSONFields | string | - | Rename fields in JSON format logs, e.g., "ts:timestamp,msg:message". |
| -loggerLevel | string | "INFO" | The minimum log level, supported: INFO, WARN, ERROR, FATAL, PANIC. |
| -loggerMaxArgLen | int | 1000 | The maximum length of a single log argument. Excess parts will be replaced with prefixes and suffixes. |
| -loggerOutput | string | "stderr" | The log output location, supported: stderr and stdout. |
| -loggerTimezone | string | "UTC" | The timezone used for log timestamps, must be a valid IANA timezone name. |
| -loggerWarnsPerSecondLimit | int | 0 | The limit on the number of warning logs per second. Excess warnings will be suppressed. 0 means no limit. |
| -memory.allowedBytes | size | 0 | The system memory size that VictoriaMetrics cache can use. If set, it overrides -memory.allowedPercent. |
| -memory.allowedPercent | float | 60 | The percentage of system memory that VictoriaMetrics cache can use. |
| -metrics.exposeMetadata | bool | false | Whether to expose TYPE and HELP metadata on the /metrics page. |
| -metricsAuthKey | value | - | The authentication key for the /metrics endpoint, supported to be read from a file or URL. |
| -pprofAuthKey | value | - | The authentication key for the /debug/pprof/* endpoints, supported to be read from a file or URL. |
| -prevCacheRemovalPercent | float | 0.1 | The percentage of service requests in the previous cache that, when below this value, will remove cache entries, improving memory usage efficiency but increasing CPU usage. |
| -pushmetrics.disableCompression | bool | false | Whether to disable request body compression when pushing metrics. |
| -pushmetrics.extraLabel | array | - | Additional labels to add when pushing metrics, e.g., instance="foo". |
| -pushmetrics.header | array | - | Additional HTTP headers to add when pushing metrics. |
| -pushmetrics.interval | duration | 10s | The interval for pushing metrics. |
| -pushmetrics.url | array | - | The list of URLs to push metrics to. |
| -queryPermitRatio | float | 1 | The ratio of query requests allowed. |
| -replicationFactor | array | 1 | The number of replicas for each sample, supports key:value format. |
| -retentionFilters.cleanInterval | duration | 360h | The interval for cleaning expired data. |
| -retentionFilters.config | string | - | The path to the retention filter configuration file. |
| -retentionFilters.configCheckInterval | duration | 1h | The interval for checking changes in the retention filter configuration file. |
| -scopeDB.dbName | string | "scopedb" | The database name for ScopeDB. |
| -scopeDB.dumpIngestBody | bool | false | Whether to print the ingestion body when it fails. |
| -scopeDB.ensureJob | bool | false | Whether to enable the ensure job feature. |
| -scopeDB.indexCacheTTL | duration | 5m | The TTL for the ScopeDB index cache. |
| -scopeDB.maxConcurrentIndexRequests | int | 32 | The maximum number of concurrent index requests. |
| -scopeDB.node | array | - | The list of ScopeDB node addresses. |
| -scopeDB.nodegroup.ctl | string | "query-meta" | The name of the control plane node group. |
| -scopeDB.nodegroup.detailQuery | string | "detail-query" | The name of the non-aggregation query node group. |
| -scopeDB.nodegroup.job | string | "job" | The name of the job execution node group. |
| -scopeDB.nodegroup.largeQuery | string | "large-query" | The name of the large query node group. |
| -scopeDB.nodegroup.query | string | "query" | The name of the regular query node group. |
| -scopeDB.nodegroup.systemQuery | string | "system-query" | The name of the system query node group from referrers like "func" and "kodo". |
| -scopeDB.nodesUpdateInterval | duration | 10s | The interval for updating nodes. |
| -scopeDB.onlyUseBootstrapNodes | bool | false | Whether to use only bootstrap nodes. |
| -scopeDB.password | string | - | The password for ScopeDB. |
| -scopeDB.port | uint | 6543 | The port for ScopeDB. |
| -scopeDB.storage.accessKeyID | string | - | The access key ID for ScopeDB storage. |
| -scopeDB.storage.bucket | string | - | The bucket for ScopeDB storage. |
| -scopeDB.storage.endpoint | string | - | The endpoint for ScopeDB storage service. |
| -scopeDB.storage.prefix | string | - | The prefix for ScopeDB storage. |
| -scopeDB.storage.region | string | - | The region for ScopeDB storage. |
| -scopeDB.storage.secretAccessKey | string | - | The secret access key for ScopeDB storage. |
| -scopeDB.storage.type | string | - | The type of ScopeDB storage. |
| -scopeDB.username | string | - | The username for ScopeDB. |
| -search.cacheTimestampOffset | duration | 5m | The maximum time offset for query response data, avoiding response gaps due to clock synchronization issues. |
| -search.compatibleMode | bool | false | Enable Prometheus compatibility mode. |
| -search.debugThrottler | bool | false | Debug the throttler. |
| -search.denyPartialResponse | bool | false | Whether to deny partial responses for data consistency. |
| -search.disableCache | bool | false | Disable response caching, suitable for backfilling historical data. |
| -search.disableImplicitConversion | bool | false | Disable implicit subquery conversion-related queries. |
| -search.dorisColdQueryCPULimit | int | 4 | The CPU limit for cold data query SQL variables. |
| -search.dorisDisableADBC | bool | false | Prefer Thrift over ADBC. |
| -search.dorisDisableSampling | bool | false | Disable sampling. |
| -search.dorisEnableSmallColdGroup | bool | false | Use a dedicated workgroup for small cold queries. |
| -search.dorisLargeColdQueryRowCount | int | - | Deprecated, use search.minRowsForLargeQuery instead. |
| -search.dorisMaxCursorRows | int | 50000000 | The maximum number of rows to scan in cursor queries. |
| -search.dorisMaxLiteralArraySize | int | 1000 | The maximum number of elements in array literals. |
| -search.dorisMaxOrderingRows | int | - | The maximum number of rows to scan in custom ordering queries. |
| -search.dorisMaxRowsForOrder | int | - | The maximum number of rows to use for ORDER BY on non-time columns. |
| -search.dorisMaxSamplingRows | int | - | Deprecated, use search.numRowsToSample instead. |
| -search.dorisThrottlerURL | string | - | The API address for the throttler. |
| -search.latencyOffset | duration | 30s | The delay after which data points become visible in query results after collection. It can be overridden by the latency_offset query parameter. A value too small may result in incomplete data points in query results. |
| -search.logImplicitConversion | bool | - | Whether to log queries that involve implicit subquery conversion. For details, see Implicit Conversion Explanation. Can be disabled with -search.disableImplicitConversion. |
| -search.logQueryMemoryUsage | size | 524288000 (500MB) | Log queries that use more than this amount of memory, facilitating the detection and optimization of large queries. Default is off. Supports KB, MB, GB, etc. |
| -search.maxConcurrentRequestsPerTenant | int | 5 | The maximum number of concurrent query requests per tenant. A value too high can cause CPU and memory resource strain. |
| -search.maxExportDuration | duration | 720h (30 days) | The maximum duration for calls to the /api/v1/export interface. |
| -search.maxGroups | int | 100000 | The maximum number of groups allowed in query results. |
| -search.maxLabelsAPIDuration | duration | 5s | The maximum response duration for /api/v1/labels, /api/v1/label/.../values, and /api/v1/series interfaces. |
| -search.maxLimit | int | 10000 | The maximum number of rows allowed in query results. |
| -search.maxLookback | duration | Dynamic detection | Synonymous with Prometheus's -search.lookback-delta. Can be overridden by the max_lookback parameter. Affects the lookback time window in queries. |
| -search.maxMemoryPerQuery | size | 0 | The maximum memory allowed for a single query. Queries exceeding this limit will be rejected. Supports KB, MB, GB, etc. |
| -search.maxMetrics | int | 10000000 | Deprecated, migrated to -search.maxShowUniqueTimeseries and -search.maxQueryUniqueTimeseries. |
| -search.maxPointsPerTimeseries | int | 30000 | The maximum number of data points returned for a single time series in /api/v1/query_range, mainly limiting the number of points for plotting clients. |
| -search.maxPointsSubqueryPerTimeseries | int | 100000 | The maximum number of data points generated per time series in subqueries. For details, see the related blog. |
| -search.maxQueryDuration | duration | 30s | The maximum execution duration for queries. |
| -search.maxQueryLen | size | 16384 (16KB) | The maximum length of a query statement. Supports KB, MB, etc. |
| -search.maxQueryUniqueTimeseries | int | 300000 | The maximum number of unique time series allowed to be selected in /api/v1/query queries. Limits memory usage. |
| -search.maxQueueDuration | duration | 10s | The maximum queuing time for requests when the maximum concurrent request limit is reached. |
| -search.maxResponseSeries | int | 0 (no limit) | The maximum number of time series returned in /api/v1/query and /api/v1/query_range. 0 means no limit. |
| -search.maxSamplesPerQuery | int | 1000000000 | The maximum number of raw samples scanned in a single query. Prevents resource exhaustion from overly large queries. |
| -search.maxSamplesPerSeries | int | 30000000 | The maximum number of raw samples scanned per time series. |
| -search.maxSeriesPerAggrFunc | int | 1000000 | The maximum number of time series generated by aggregation functions. |
| -search.maxShowUniqueTimeseries | int | 10000000 | The maximum number of time series allowed to be displayed by show functions. Limits memory usage. |
| -search.maxStalenessInterval | duration | Dynamic calculation | The maximum time interval for staleness calculation, automatically calculated based on the median interval of samples by default. |
| -search.maxStatusRequestDuration | duration | 5m0s | The maximum response time for /api/v1/status/* requests. |
| -search.maxStepForPointsAdjustment | duration | 1m0s | The maximum step for adjusting points in /api/v1/query_range when the distance to the current time is less than latencyOffset. |
| -search.maxWorkersPerQuery | int | 2 | The maximum number of CPU cores allowed for a single query. Can be adjusted to optimize performance, but cannot exceed the number of physical CPU cores. |
| -search.minRowsForLargeQuery | int | 10000000 | The minimum number of rows to scan for large queries. |
| -search.minStalenessInterval | duration | - | The minimum time interval for staleness calculation, used to remove chart gaps caused by irregular sampling intervals. |
| -search.minTimeRangeForLargeQuery | duration | 12h0m0s | The minimum time range for large queries. |
| -search.minWindowForInstantRollupOptimization | duration | 3h | The minimum window value for enabling cache optimization for instant queries with lookback windows. |
| -search.noStaleMarkers | bool | false | If the database has no Prometheus staleness markers, setting this to true can save CPU resources. |
| -search.queryStats.lastQueriesCount | int | 20000 | The number of recent queries tracked by /api/v1/status/top_queries. 0 disables tracking. |
| -search.queryStats.minQueryDuration | duration | 1ms | The minimum query duration threshold for statistics. Queries shorter than this are ignored. |
| -search.resetRollupResultCacheOnStartup | bool | false | Whether to reset the rollup result cache on startup. |
| -search.rowsToSample | int | 10000000 | The maximum number of rows to scan for sampling queries. |
| -search.rowsToSampleForFacet | int | 10000000 | The maximum number of rows to scan for show_x_field or distinct queries. |
| -search.scopeDB.firstCursorScanRows | int | 10000000 | The maximum number of rows to scan in the first cursor pagination query. |
| -search.scopeDB.logQuery | bool | true | Whether to log ScopeDB query statements. |
| -search.scopeDB.maxLiteralArraySize | int | 1000 | The maximum number of elements in array literals. |
| -search.scopeDB.maxRowsForOrder | int | 100000 | The maximum number of rows for ORDER BY on non-time columns. |
| -search.scopeDB.maxTimeRange | duration | - | The maximum time range for queries. |
| -search.setLookbackToStep | bool | false | Whether to fix lookback to the step parameter, adjusting the query model to be more like InfluxDB. When enabled, -search.maxLookback and -search.maxStalenessInterval are ignored. |
| -search.skipSlowReplicas | bool | false | Whether to skip the slowest vmstorage node in the replication factor during queries, potentially improving query speed but risking incomplete data. |
| -search.treatDotsAsIsInRegexps | bool | false | Whether to filter dot characters in regular expressions. This option is deprecated; it is recommended to use graphite syntax. |
| -slowQueryDuration | duration | 1s | Queries that exceed this duration will be logged as slow queries. |
| -storageNode | array | - | The array of guance-storage node addresses, comma-separated or specified multiple times. |
| -streamaggr.cleanupInterval | duration | 10m0s | The interval for cleaning up expired queries. |
| -streamaggr.dataBlocksCacheSize | int | 157286400 (150MB) | The size of the data block cache. |
| -streamaggr.ignoreContinuousCheck | bool | false | Whether to ignore continuity checks. |
| -streamaggr.indexBlocksCacheSize | int | 52428800 (50MB) | The size of the index block cache. |
| -streamaggr.queryRetentionPeriod | duration | 168h0m0s (7 days) | The retention period for queries. |
| -streamaggr.retentionMonths | int | 6 | The number of months to retain stream aggregation data. |
| -streamaggr.selectVerify | bool | false | Whether to verify stream aggregation results. |
| -tls | array | None | Whether to enable TLS (i.e., HTTPS) for the specified -httpListenAddr. When setting this parameter, -tlsCertFile and -tlsKeyFile must also be set. Supports multiple values separated by commas or specified multiple times. An empty value is treated as false. |
| -tlsCertFile | array | None | The path to the TLS certificate file for the corresponding -httpListenAddr. It is recommended to use ECDSA certificates instead of RSA certificates, as RSA certificates are slower. The certificate file is automatically re-read every second, allowing for dynamic updates. Supports multiple values. |
| -tlsCipherSuites | array | None | An optional list of TLS cipher suites to use for HTTPS requests when TLS is enabled. Supports multiple values. For a detailed list of supported cipher suites, see: https://pkg.go.dev/crypto/tls#pkg-constants |
| -tlsKeyFile | array | None | The path to the TLS key file for the corresponding -httpListenAddr. The key file is automatically re-read every second, allowing for dynamic updates. Supports multiple values. |
| -tlsMinVersion | array | None | The minimum TLS version to allow (optional). Supports TLS10, TLS11, TLS12, TLS13. Supports multiple values. |
| -version | No parameter | None | Display the version information of VictoriaMetrics. |
| -vmstorageDialTimeout | duration | 3s | The timeout for establishing an RPC connection between vmselect and vmstorage. See -vmstorageUserTimeout for details. |
| -vmstorageUserTimeout | duration | 3s | The network timeout for RPC connections between vmselect and vmstorage (Linux only). A lower value reduces the maximum query duration in case of network failures. See TCP_USER_TIMEOUT for details. |
| -zipkin.batchSize | int | 100 | The maximum number of Spans to send in a single batch. |
| -zipkin.bufferSize | int | 10000 | The maximum number of Spans to buffer. |
| -zipkin.endpoint | string | None | The endpoint URL of the Zipkin collector, e.g., http://localhost:9529. |
| -zipkin.flushPeriod | duration | 1s | The maximum interval between flushes. |
| -zipkin.serviceName | string | guance-select | The service name used in Zipkin Spans. |
Guance Insert Parameter Description¶
| Parameter Name | Type | Default Value | Description |
|---|---|---|---|
| -blockcache.missesBeforeCaching | int | 2 | The number of cache misses before a cache block is placed into the cache. A higher value may reduce cache size but increase CPU and disk read pressure. |
| -cacheExpireDuration | duration | 30m0s | Data in the memory cache is removed after being unaccessed for a certain period. A smaller value may reduce memory usage but increase CPU usage. |
| -denyQueryTracing | None | None | Whether to disable query tracing. See: https://docs.victoriametrics.com/#query-tracing |
| -disableInflux | None | true | Whether to disable support for Influx format. |
| -disableRerouting | None | true | Whether to disable data rerouting. Rerouting can be triggered when some vmstorage nodes write slowly. Turning off this feature can reduce active time series during rolling restarts. |
| -disableReroutingOnUnavailable | None | None | Whether to disable rerouting when vmstorage nodes are unavailable. When disabled, data writing stops if a node is unavailable, reducing the impact during restarts or drastic sequence changes. |
| -doris.cacheDataPath | string | /data/file-queue | The path to the cache file. |
| -doris.coldDataTierOnly | None | None | Use only the cold data storage tier. |
| -doris.debugIndexName | string | default | The indexName to add debug logs for. |
| -doris.debugMeasurement | string | None | The measurement to add debug logs for. |
| -doris.debugNamespace | string | None | The namespace to add debug logs for. |
| -doris.debugTenantID | uint | None | The tenant ID (accountID) to add debug logs for. |
| -doris.disableBFOnNewTables | None | None | Do not enable Bloom filters for new tables. |
| -doris.disableKeepAlives | None | None | Disable keep-alive in stream load. |
| -doris.disableUnicodeTokenizer | None | None | Use a Chinese tokenizer instead of a Unicode tokenizer. |
| -doris.enableBFOnOldTables | None | None | Enable Bloom filters for old tables. |
| -doris.enableFileCacheTTL | None | None | Enable cloud database SelectDB data cache: https://help.aliyun.com/document_detail/2638759.html |
| -doris.enableWarmDataTier | None | None | Enable hot + warm storage mode instead of hot + cold mode. |
| -doris.forceFlushInterval | duration | 30s | The forced flush interval for non-event data. |
| -doris.indexCacheTTL | duration | 30m0s | The TTL for the Doris index cache. |
| -doris.loadToSingleTablet | None | None | Use the load_to_single_tablet header in stream load requests. |
| -doris.maxCacheSize | size | 0 | Deprecated, use fileQueue.maxTotalSize instead. The maximum number of bytes for total cached data. |
| -doris.maxConcurrentInserts | int | None | Deprecated, use fileQueue.workers instead. |
| -doris.maxFieldCount | size | 512 | The maximum number of fields allowed per measurement. Supports KB, MB, GB, etc. |
| -doris.maxFieldSize | size | 33554432 | The maximum byte size for a single field. Supports KB, MB, GB, etc. |
| -doris.maxSizePerRequest | size | 0 | Deprecated, use fileQueue.maxFileSize instead. The maximum size of a single line in a single parse. |
| -doris.onlyQueryFE | None | None | When enabled, query data only from FE nodes, not directly from BE. |
| -doris.onlyWriteFE | None | None | When enabled, write data only to FE nodes, not directly to BE. |
| -doris.replicationFactor | int | 1 | The replication factor for Doris tables. |
| -doris.sampleIndexName | string | None | The sample index name. |
| -doris.sampleInsertRatio | int | None | The sample insert ratio. |
| -doris.sampleInterval | value | 1s | The sample interval time, supports s/m/h/d/w/y as units. Defaults to monthly. |
| -doris.sampleNamespace | string | None | The sample namespace. |
| -doris.sampleTenantID | string | None | The sample tenant ID. |
| -doris.storageResource | string | s3_default | The name of the storage policy resource used by Doris. |
| -doris.variantSchemaMode | string | None | The schema mode, supports 'none', 'minimal', 'full'. |
| -dorisCluster | string | default_cluster | The name of the Doris cluster. |
| -dorisFENode | array | None | The Doris-FE node addresses (multiple values separated by commas). |
| -dorisHTTPPort | int | 8030 | The HTTP port of Doris. |
| -dorisMySQLPort | int | 9030 | The MySQL port of Doris. |
| -dorisPassword | string | None | The password for connecting to Doris. |
| -dorisUsername | string | root | The username for connecting to Doris. |
| -dropSamplesOnOverload | None | None | Whether to discard samples if vmstorage nodes are overloaded or unavailable to ensure cluster availability. Note that this operation is performed before replication, so use with caution. |
| -enableTCP6 | None | None | Whether to enable IPv6 listening and dialing. By default, only IPv4 is enabled. |
| -envflag.enable | None | None | Whether to allow reading parameters from environment variables. Command-line parameters take precedence. See: https://docs.victoriametrics.com/#environment-variables |
| -envflag.prefix | string | None | The prefix for environment variables (requires -envflag.enable to be enabled). |
| -fileQueue.maxFileSize | size | 104857600 | The maximum size of a single file in the file queue (default 100MB). |
| -fileQueue.maxTotalSize | size | 3758096384 | The total size limit of the file queue (default 3.5GB). |
| -fileQueue.workers | int | 8 | The maximum number of concurrent threads for inserting into ScopeDB or Doris. |
| -filedb.maxFileSize | int | 10485760 | The maximum file size for filedb (default 10MB). |
| -filestream.disableFadvise | None | None | Whether to disable the fadvise() system call, which prevents recently accessed data from being evicted from the OS cache when reading large files. |
| -flagsAuthKey | value | None | The authentication key for the /flags interface. Can be read from a file or URL. |
| -fs.disableMmap | None | None | Whether to disable mmap() and use pread() to read data files. By default, mmap() is enabled on 64-bit systems. |
| -gcolumnCompressLevel | int | None | The compression level for the guance column protocol. Higher values compress better but consume more CPU; negative values reduce CPU usage but increase network transmission. |
| -http.connTimeout | duration | 2m0s | The timeout for HTTP connections. Can be used to make load balancers distribute traffic more evenly. 0 means no timeout. |
| -http.disableResponseCompression | None | None | Whether to disable HTTP response compression to save CPU. By default, compression is enabled to save network bandwidth. |
| -http.header.csp | string | None | Set the Content-Security-Policy header value, recommended value: "default-src 'self'". |
| -http.header.frameOptions | string | None | Set the X-Frame-Options header value. |
| -http.header.hsts | string | None | Set the Strict-Transport-Security header value, recommended value: "max-age=31536000; includeSubDomains". |
| -http.idleConnTimeout | duration | 1m0s | The timeout for idle HTTP connections. |
| -http.maxGracefulShutdownDuration | duration | 7s | The maximum wait time for graceful shutdown of the HTTP service. High-load servers may need a higher value. |
| -http.pathPrefix | string | None | The prefix for all HTTP paths, e.g., setting to /foo/bar will make all paths start with this, facilitating reverse proxy. |
| -http.shutdownDelay | duration | None | The delay before the HTTP service shuts down. During this time, the /health interface returns non-OK, notifying load balancers to switch traffic. |
| -httpAuth.password | value | None | The password for HTTP Basic Auth. Authentication is disabled if the username is empty. Can be loaded from a file or URL. |
| -httpAuth.username | string | None | The username for HTTP Basic Auth. If empty, authentication is disabled. Requires -httpAuth.password to be set. |
| -httpListenAddr | array | None | The HTTP service listening address. Supports multiple addresses. |
| -httpListenAddr.useProxyProtocol | array | false | Whether to use proxy protocol for connections received on -httpListenAddr. When enabled, the HTTP server cannot provide the /metrics endpoint. Use -pushmetrics.url for metric pushing. Supports multiple values separated by commas or specified multiple times. An empty value is treated as false. |
| -influx.defaultBatchLinesCount | size | 16 | The default number of lines per batch for InfluxDB parsing. Supports unit suffixes: KB, MB, GB, TB, KiB, MiB, GiB, TiB. |
| -influx.defaultBatchSize | size | 1048576 | The default byte size per batch for InfluxDB parsing. Supports unit suffixes: KB, MB, GB, TB, KiB, MiB, GiB, TiB. |
| -insert.maxFutureTimeDrift | duration | 1h0m0s | Events with timestamps greater than the current time plus this value will be rejected during insertion. |
| -insert.maxPastTimeDrift | duration | Events with timestamps less than the current time minus this value will be rejected during insertion. | |
| -insertNode | array | The addresses of guance-insert nodes, multiple values separated by commas, supports values with quotes or brackets. | |
| -internStringCacheExpireDuration | duration | 6m0s | The expiration duration for interned string cache. See: https://en.wikipedia.org/wiki/String_interning. |
| -internStringDisableCache | false | Whether to disable the interned string cache. Disabling it may reduce memory usage but increase CPU usage. | |
| -internStringMaxLen | int | 500 | The maximum length of interned strings. Setting a lower value may reduce memory usage but increase CPU usage. |
| -loggerDisableTimestamps | false | Whether to disable timestamp writing in logs. | |
| -loggerErrorsPerSecondLimit | int | The maximum number of ERROR logs allowed per second. Excess errors will be suppressed. 0 means no limit. | |
| -loggerFormat | string | default | The log format, options: default, json. |
| -loggerJSONFields | string | - | Rename fields in JSON format logs, e.g., "ts:timestamp,msg:message". Supported fields: ts, level, caller, msg. |
| -loggerLevel | string | INFO | The minimum log level, supported: INFO, WARN, ERROR, FATAL, PANIC. |
| -loggerMaxArgLen | int | 1000 | The maximum length of a single log argument. Excess parts will be replaced with prefixes and suffixes. |
| -loggerOutput | string | stderr | The log output location, supported: stderr, stdout. |
| -loggerTimezone | string | UTC | The timezone used for log timestamps, must be a valid IANA timezone name, e.g., America/New_York, Etc/GMT+3, or Local. |
| -loggerWarnsPerSecondLimit | int | The maximum number of WARN logs allowed per second. Excess warnings will be suppressed. 0 means no limit. | |
| -maxConcurrentInserts | int | 24 | The maximum number of concurrent insert requests allowed. The default value is suitable for most cases and can be increased when the network is slow. |
| -maxInsertRequestSize | size | 134217728 | The maximum byte size for a single gcolumn insert request. Supports unit suffixes: KB, MB, GB, TB, KiB, MiB, GiB, TiB. |
| -maxLabelValueLen | int | 16384 | The maximum length of label values in received time series. Excess parts will be truncated, and the vm_too_long_label_values_total metric on the /metrics page will increase. |
| -maxLabelsPerTimeseries | int | 256 | The maximum number of labels allowed per time series. Excess labels will be discarded, and the vm_metrics_with_dropped_labels_total metric on the /metrics page will increase. |
| -memory.allowedBytes | size | 0 | The maximum system memory that VictoriaMetrics cache can occupy (in bytes). If non-zero, it overrides -memory.allowedPercent. Setting it too small increases cache misses, and setting it too large may affect the OS page cache. |
| -memory.allowedPercent | float | 60 | The percentage of system memory that VictoriaMetrics cache can occupy. Setting it too small may increase CPU and disk IO, and setting it too large may affect the page cache. |
| -metrics.exposeMetadata | false | Whether to expose TYPE and HELP metadata on the /metrics page. Some systems (e.g., Google Cloud Managed Prometheus) require this information. | |
| -metricsAuthKey | value | The authentication key for the /metrics endpoint, can be passed via the authKey query parameter. Supports reading the key from a file or URL. | |
| -pprofAuthKey | value | The authentication key for the /debug/pprof/* endpoints, can be passed via the authKey query parameter. Supports reading the key from a file or URL. | |
| -prevCacheRemovalPercent | float | 0.1 | The percentage of old cache hits below which entries are removed. Higher values reduce memory usage but may increase CPU usage. |
| -pushmetrics.disableCompression | false | Whether to disable request body compression when pushing metrics to each -pushmetrics.url. | |
| -pushmetrics.extraLabel | array | Extra labels to add when pushing metrics to each -pushmetrics.url, e.g., -pushmetrics.extraLabel='instance="foo"'. Supports commas, quotes, and brackets. |
|
| -pushmetrics.header | array | HTTP Headers to include when pushing metrics to each -pushmetrics.url, e.g., Authorization: Basic foobar. Supports commas, quotes, and brackets. |
|
| -pushmetrics.interval | duration | 10s | The interval for pushing metrics to each -pushmetrics.url. |
| -pushmetrics.url | array | The target addresses to push /metrics page data to. By default, no pushing is done. Supports multiple addresses and complex syntax. | |
| -relabelConfig | string | The path to the relabeling rules configuration file (local or http address). Automatically reloaded upon receiving SIGHUP. See: https://docs.victoriametrics.com/#relabeling | |
| -replicationFactor | int | 1 | The number of replicas for data, i.e., the number of copies of each data on different -storageNode. If this value >1, vmselect must set -dedup.minScrapeInterval=1ms for deduplication. |
| -retentionFilters.cleanInterval | duration | 360h0m0s | The interval for cleaning expired data. |
| -retentionFilters.config | string | The path to the retention policy configuration file. | |
| -retentionFilters.configCheckInterval | duration | 1h0m0s | The interval for checking changes in the retention policy configuration file. |
| -rpc.disableCompression | false | Whether to disable compression when sending data from vminsert to vmstorage. Disabling it reduces CPU usage but increases network bandwidth usage. | |
| -scopeDB.dbName | string | scopedb | The database name for ScopeDB. |
| -scopeDB.dumpIngestBody | false | Whether to output the request body content when insertion fails. | |
| -scopeDB.ensureJob | false | Whether to enable ensure job. | |
| -scopeDB.indexCacheTTL | duration | 5m0s | The TTL for the ScopeDB index cache. |
| -scopeDB.maxConcurrentIndexRequests | int | 32 | The maximum number of concurrent index requests allowed for ScopeDB. |
| -scopeDB.maxFieldCount | size | 512 | The maximum number of fields per index. |
| -scopeDB.maxFieldSize | size | 33554432 | The maximum byte size for a single gcolumn field. |
| -scopeDB.node | array | The list of ScopeDB node addresses, supports multiple formats. | |
| -scopeDB.nodegroup.ctl | string | ingest-meta | The ScopeDB node group used by the control plane. |
| -scopeDB.nodegroup.ingest | string | ingest | The ScopeDB node group used for data ingestion. |
| -scopeDB.nodesUpdateInterval | duration | 10s | The interval for updating ScopeDB node information. |
| -scopeDB.onlyUseBootstrapNodes | bool | false | Whether to use only bootstrap nodes. |
| -scopeDB.password | string | The access password for ScopeDB. | |
| -scopeDB.port | uint | 6543 | The access port for ScopeDB. |
| -scopeDB.username | string | The username for ScopeDB. | |
| -search.debugThrottler | bool | false | Debug the throttler. |
| -search.denyPartialResponse | bool | false | Whether to deny partial responses (improves consistency at the cost of availability). |
| -search.dorisColdQueryCPULimit | int | 4 | The CPU limit for cold data queries in Doris. |
| -search.dorisDisableADBC | bool | false | Whether to disable ADBC (prefer Thrift). |
| -search.dorisDisableSampling | bool | false | Whether to disable sampling. |
| -search.dorisEnableSmallColdGroup | bool | false | Enable a dedicated workgroup for small cold queries. |
| -search.dorisMaxCursorRows | int | 50000000 | The maximum number of rows to scan in cursor queries. |
| -search.dorisMaxLiteralArraySize | int | 1000 | The maximum number of elements in literal value arrays. |
| -search.dorisMaxOrderingRows | int | The maximum number of rows to scan in custom ordering queries. | |
| -search.dorisMaxRowsForOrder | int | The maximum number of rows for ORDER BY on non-time columns. | |
| -search.dorisThrottlerURL | string | The API address for the Throttler. | |
| -search.logQueryMemoryUsage | size | 524288000 | Log queries that use more than the specified amount of memory. |
| -search.maxExportDuration | duration | 720h0m0s | The maximum duration for /api/v1/export requests. |
| -search.maxGroups | int | 100000 | The maximum number of groups allowed in query results. |
| -search.maxLabelsAPIDuration | duration | 5s | The maximum response duration for /api/v1/labels, /api/v1/label/.../values, and /api/v1/series requests. |
| -search.maxLimit | int | 10000 | The maximum number of rows allowed in query results. |
| -search.maxMemoryPerQuery | size | 0 | The maximum memory allowed for a single query. Queries exceeding this limit will be rejected. Supports KB, MB, GB, etc. |
| -search.maxQueryDuration | duration | 30s | The maximum execution duration for queries. |
| -search.maxStatusRequestDuration | duration | 5m0s | The maximum response duration for /api/v1/status/* requests. |
| -search.minTimeRangeForLargeQuery | duration | 12h0m0s | The minimum time range for large queries. |
| -search.scopeDB.firstCursorScanRows | int | 10000000 | The number of rows to scan in the first cursor pagination query. |
| -search.scopeDB.logQuery | bool | true | Whether to log ScopeDB query statements. |
| -search.scopeDB.maxLiteralArraySize | int | 1000 | The maximum number of elements in literal value arrays. |
| -search.scopeDB.maxRowsForOrder | int | 100000 | The maximum number of rows for ORDER BY on non-time columns. |
| -search.scopeDB.maxTimeRange | duration | The maximum time range for queries. | |
| -sortLabels | bool | false | Whether to sort labels before writing to storage, reducing memory usage but potentially decreasing write performance. |
| -storageNode | array | The list of guance-storage node addresses, supports multiple formats. | |
| -streamaggr.cleanupInterval | duration | 10m0s | The interval for cleaning up expired queries. |
| -streamaggr.dataBlocksCacheSize | int | 157286400 | The size of the data block cache (in bytes). |
| -streamaggr.ignoreContinuousCheck | bool | false | Whether to ignore continuous checks. |
| -streamaggr.indexBlocksCacheSize | int | 52428800 | The size of the index block cache (in bytes). |
| -streamaggr.insertVerify | bool | false | Enable streamaggr verification mode. |
| -streamaggr.queryRetentionPeriod | duration | 168h0m0s | The retention period for queries. |
| -streamaggr.replay | bool | false | Enable streamaggr replay mode. |
| -streamaggr.retentionMonths | int | 6 | The number of months to retain stream aggregation data. |
| -streamaggrDataPath | string | The path to store data. | |
| -tls | array | Whether to enable TLS for the specified httpListenAddr. Requires setting -tlsCertFile and -tlsKeyFile. | |
| -tlsCertFile | array | The path to the TLS certificate file, supports dynamic updates. | |
| -tlsCipherSuites | array | An optional list of TLS cipher suites. | |
| -tlsKeyFile | array | The path to the TLS key file, supports dynamic updates. | |
| -tlsMinVersion | array | The minimum TLS version (TLS10, TLS11, TLS12, TLS13). | |
| -usePromCompatibleNaming | bool | false | Whether to replace unsupported Prometheus characters with underscores. |
| -version | bool | false | Display the version information of VictoriaMetrics. |
| -vmstorageDialTimeout | duration | 3s | The timeout for establishing an RPC connection from vminsert to vmstorage. |
| -vmstorageUserTimeout | duration | 3s | The network timeout for connections from vminsert to vmstorage (Linux only). |
| -zipkin.batchSize | int | 100 | The maximum number of spans to send in a single batch. |
| -zipkin.bufferSize | int | 10000 | The maximum number of spans to buffer. |
| -zipkin.endpoint | string | The endpoint URL of the Zipkin collector. | |
| -zipkin.flushPeriod | duration | 1s | The maximum interval between flushes. |
| -zipkin.serviceName | string | guance-insert | The service name used in Zipkin Spans. |
Guance Storage Parameter Description¶
| Parameter Name | Type | Default Value | Description |
|---|---|---|---|
| -bigMergeConcurrency | int | None | Deprecated: This parameter no longer has any effect. |
| -blockcache.missesBeforeCaching | int | 2 | The number of cache misses before a cache block is placed into the cache. A higher value may reduce the size of the indexdb/dataBlocks cache but increase CPU and disk read operations. |
| -cacheExpireDuration | duration | 30m0s | Unaccessed entries in the memory cache are removed after this duration. A lower value may reduce memory usage but increase CPU usage. See -prevCacheRemovalPercent. |
| -dedup.minScrapeInterval | duration | None | Only the last sample within each discrete interval is retained for each time series. Must be set greater than 0. See deduplication documentation. |
| -denyQueriesOutsideRetention | bool | false | Whether to deny queries for data outside the configured -retentionPeriod. When enabled, /api/v1/query_range returns 503 for queries outside the retention period. |
| -denyQueryTracing | bool | false | Whether to disable query tracing. See query tracing documentation. |
| -enableTCP6 | bool | false | Whether to enable IPv6 for listening and dialing. By default, only IPv4 is enabled. |
| -envflag.enable | bool | false | Whether to enable reading parameters from environment variables, with command-line parameters taking precedence. If not enabled, parameters are only read from the command line. See environment variable documentation. |
| -envflag.prefix | string | None | The prefix for environment variables when -envflag.enable is enabled. |
| -filestream.disableFadvise | bool | false | Whether to disable the fadvise() system call when reading large files. This call prevents recently accessed data from being evicted from the system cache. Disabling it may reduce CPU usage in some cases. |
| -finalMergeDelay | duration | None | Deprecated: This parameter no longer has any effect. |
| -flagsAuthKey | value | None | The authentication key for the /flags endpoint, must be passed via the authKey query parameter. Supports reading the key value from a file or HTTP(S) URL. |
| -forceFlushAuthKey | value | None | The authentication key required for the /internal/force_flush page, supports reading from a file or HTTP(S) URL. |
| -forceMergeAuthKey | value | None | The authentication key required for the /internal/force_merge page, supports reading from a file or HTTP(S) URL. |
| -fs.disableMmap | bool | false | Whether to use pread() instead of mmap() to read data files. By default, mmap() is used on 64-bit architectures and pread() is used on 32-bit architectures. mmap() is faster for reading small data blocks. |
| -http.connTimeout | duration | 2m0s | The timeout for closing connections to -httpListenAddr. Helps load balancers distribute requests evenly. 0 disables timeout. |
| -http.disableResponseCompression | bool | false | Whether to disable HTTP response compression. By default, compression is enabled to save network bandwidth. Disabling it saves CPU. |
| -http.header.csp | string | "default-src 'self'" | Set the HTTP header Content-Security-Policy, recommended value is "default-src 'self'". |
| -http.header.frameOptions | string | None | Set the HTTP header X-Frame-Options. |
| -http.header.hsts | string | "max-age=31536000; includeSubDomains" | Set the HTTP header Strict-Transport-Security, recommended value as shown. |
| -http.idleConnTimeout | duration | 1m0s | The timeout for closing idle HTTP connections. |
| -http.maxGracefulShutdownDuration | duration | 7s | The maximum wait time for graceful shutdown of the HTTP server. High-load servers may need to increase this value. |
| -http.pathPrefix | string | None | Add a prefix to all HTTP paths, e.g., for proxy path prefixes. See related documentation. |
| -http.shutdownDelay | duration | None | The delay before the HTTP server shuts down. During this time, /health returns non-OK status, making it easier for load balancers to switch requests. |
| -httpAuth.password | value | None | The password for HTTP basic authentication. Authentication is disabled if the username is not set. Supports reading from a file or HTTP(S) URL. |
| -httpAuth.username | string | None | The username for HTTP basic authentication. If not set, authentication is disabled. |
| -httpListenAddr | array | None | The HTTP listening address. Supports multiple values or complex formats. |
| -httpListenAddr.useProxyProtocol | array | false | Whether to enable Proxy Protocol for the specified listening address. When enabled, the regular /metrics endpoint cannot be used. See HAProxyProxy documentation. |
| -inmemoryDataFlushInterval | duration | 5s | The interval for flushing in-memory data to disk, saving data in case of abnormal shutdown. A larger interval increases flash lifespan, while a smaller interval increases disk IO load. Minimum 1 second. |
| -insert.maxQueueDuration | duration | 1m0s | The maximum queuing time for insert requests when the maximum concurrent inserts are reached. |
| -internStringCacheExpireDuration | duration | 6m0s | The TTL for the interned string cache. See string interning mechanism. |
| -internStringDisableCache | bool | false | Whether to disable the interned string cache. Disabling it reduces memory but increases CPU usage. |
| -internStringMaxLen | int | 500 | The maximum length of interned strings. A lower value saves memory but increases CPU usage. |
| -logNewSeries | bool | false | Whether to log new time series. Only used for debugging, may affect performance. |
| -loggerDisableTimestamps | bool | None | Whether to disable timestamps in logs. |
| -loggerErrorsPerSecondLimit | int | 0 | The maximum number of ERROR logs allowed per second. Excess logs will be suppressed. 0 means no limit. |
| -loggerFormat | string | "default" | The log format. Options: default, json (default "default"). |
| -loggerJSONFields | string | None | Rename fields in JSON format logs. For example, "ts:timestamp,msg:message" will rename "ts" to "timestamp" and "msg" to "message". Supported fields: ts, level, caller, msg. |
| -loggerLevel | string | "INFO" | The minimum log level. Options: INFO, WARN, ERROR, FATAL, PANIC (default "INFO"). |
| -loggerMaxArgLen | int | 1000 | The maximum length of a single log argument. Arguments exceeding this length will be truncated and displayed as 'arg_start..arg_end', with the prefix and suffix lengths not exceeding -loggerMaxArgLen / 2. |
| -loggerOutput | string | "stderr" | The log output location. Options: stderr, stdout (default "stderr"). |
| -loggerTimezone | string | "UTC" | The timezone used for log timestamps. The timezone must be a valid IANA timezone name, such as America/New_York, Europe/Berlin, Etc/GMT+3, or Local (default "UTC"). |
| -loggerWarnsPerSecondLimit | int | 0 | The maximum number of WARN logs allowed per second. Excess logs will be suppressed. 0 means no limit. |
| -maxConcurrentInserts | int | 4 | The maximum number of concurrent insert requests. Can be increased for slow client networks. The default value is set based on the number of CPU cores and is generally suitable. See -insert.maxQueueDuration. |
| -memory.allowedBytes | size | 0 | The maximum system memory that VictoriaMetrics cache can occupy (in bytes). If set to a non-zero value, it overrides -memory.allowedPercent. A value too low increases cache misses, leading to higher CPU and disk IO. Supports unit suffixes KB, MB, GB, etc. |
| -memory.allowedPercent | float | 60 | The percentage of system memory that VictoriaMetrics cache can occupy. See -memory.allowedBytes. A value too low increases cache misses, while a value too high may lead to more disk IO. |
| -metrics.exposeMetadata | bool | None | Whether to expose TYPE and HELP metadata on the /metrics page. Some systems (e.g., Google Cloud Managed Prometheus) require this information. |
| -metricsAuthKey | string | None | The authentication key for the /metrics endpoint. Passed via the authKey query parameter, takes precedence over httpAuth.* settings. Can be read from a file or HTTP/HTTPS URL. |
| -pprofAuthKey | string | None | The authentication key for the /debug/pprof/* endpoints. Passed via the authKey query parameter, takes precedence over httpAuth.* settings. Can be read from a file or HTTP/HTTPS URL. |
| -precisionBits | int | 64 | The number of precision bits stored per value. Fewer bits increase compression but reduce precision. (Default 64) |
| -prevCacheRemovalPercent | float | 0.1 | The percentage of service requests in the previous cache that, when below this value, will remove cache entries. Increasing this value reduces memory usage but increases CPU usage. See -cacheExpireDuration. |
| -pushmetrics.disableCompression | bool | None | Whether to disable request body compression when pushing metrics to all -pushmetrics.url. |
| -pushmetrics.extraLabel | array | None | Extra labels to add when pushing metrics to all -pushmetrics.url. For example, -pushmetrics.extraLabel='instance="foo"'. Supports arrays and multiple settings. |
| -pushmetrics.header | array | None | HTTP headers to add when pushing metrics to all -pushmetrics.url. For example, -pushmetrics.header='Authorization: Basic foobar'. Supports arrays and multiple settings. |
| -pushmetrics.interval | duration | 10s | The interval for pushing metrics to all -pushmetrics.url. |
| -loggerDisableTimestamps | bool | false | Whether to disable timestamps in logs. |
| -loggerErrorsPerSecondLimit | int | 0 | The maximum number of ERROR messages allowed per second. Excess logs will be suppressed. 0 means no limit. |
| -loggerFormat | string | "default" | The log format. Options: default, json. |
| -loggerJSONFields | string | "" | Rename fields in JSON format logs. For example, "ts:timestamp,msg:message" renames "ts" to "timestamp". Supported fields: ts, level, caller, msg. |
| -loggerLevel | string | "INFO" | The minimum log level. Options: INFO, WARN, ERROR, FATAL, PANIC. |
| -loggerMaxArgLen | int | 1000 | The maximum length of a single log argument. Arguments exceeding this length will be truncated as 'arg_start..arg_end'. |
| -loggerOutput | string | "stderr" | The log output location. Options: stderr, stdout. |
| -loggerTimezone | string | "UTC" | The timezone used for log timestamps. Must be a valid IANA timezone name, such as America/New_York. |
| -loggerWarnsPerSecondLimit | int | 0 | The maximum number of WARN messages allowed per second. Excess logs will be suppressed. 0 means no limit. |
| -maxConcurrentInserts | int | 4 | The maximum number of concurrent write requests. The default is set automatically based on the number of CPU cores. |
| -memory.allowedBytes | size | 0 | The maximum system memory that VictoriaMetrics cache can use. Takes precedence over -memory.allowedPercent. Supports units KB, MB, GB, etc. |
| -memory.allowedPercent | float | 60 | The percentage of system memory that VictoriaMetrics cache can use. |
| -metrics.exposeMetadata | bool | false | Whether to expose TYPE and HELP metadata on the /metrics page. |
| -metricsAuthKey | string | "" | The authentication key for the /metrics endpoint. Supports reading from a file or URL. |
| -pprofAuthKey | string | "" | The authentication key for the /debug/pprof/* endpoints. Supports reading from a file or URL. |
| -precisionBits | int | 64 | The number of precision bits stored per value. A lower value increases compression but reduces precision. |
| -prevCacheRemovalPercent | float | 0.1 | The percentage threshold for removing entries from the previous cache. A higher value reduces memory usage but increases CPU usage. |
| -pushmetrics.disableCompression | bool | false | Whether to disable request body compression when pushing metrics. |
| -pushmetrics.extraLabel | array | [] | Extra labels to add when pushing metrics, e.g., instance="foo". |
| -pushmetrics.header | array | [] | HTTP headers to add when pushing metrics, e.g., Authorization. |
| -pushmetrics.interval | duration | 10s | The interval for pushing metrics. |
| -pushmetrics.url | array | [] | The list of remote URLs to push metrics to. By default, no pushing is done. |
| -retentionFilters.cleanInterval | duration | 360h | The interval for cleaning expired data. |
| -retentionFilters.config | string | "" | The path to the retention filter configuration file. |
| -retentionFilters.configCheckInterval | duration | 1h | The interval for checking changes in the retention filter configuration file. |
| -retentionPeriod | value | 1 | The data retention period. Data older than this period is automatically deleted. Supports units s, m, h, d, w, y. |
| -retentionTimezoneOffset | duration | 0 | The timezone offset for indexdb rotation. 0 means UTC 4 AM. |
| -rpc.disableCompression | bool | false | Whether to disable compression between vmstorage and vmselect. Disabling it reduces CPU usage but increases network bandwidth. |
| -search.maxConcurrentRequests | int | 4 | The maximum number of concurrent vmselect requests that vmstorage can handle. |
| -search.maxQueueDuration | duration | 10s | The maximum queuing time for requests. |
| -search.maxTagKeys | int | 100000 | The maximum number of tag keys returned in a single search. |
| -search.maxTagValues | int | 100000 | The maximum number of tag values returned in a single search. |
| -search.maxUniqueTimeseries | int | 0 | The maximum number of unique time series that can be scanned in a query. 0 means no limit. |
| -smallMergeConcurrency | int | (Deprecated) | Deprecated, no longer has any effect. |
| -snapshotAuthKey | string | "" | The authentication key for the /snapshot* pages. Supports reading from a file or URL. |
| -snapshotCreateTimeout | duration | (Deprecated) | Deprecated, no longer has any effect. |
| -snapshotsMaxAge | value | 0 | Automatically delete snapshots older than this age. Supports units s, m, h, d, w, y. |
| -storage.cacheSizeIndexDBDataBlocks | size | 0 | The maximum size of the indexdb/dataBlocks cache. |
| -storage.cacheSizeIndexDBIndexBlocks | size | 0 | The maximum size of the indexdb/indexBlocks cache. |
| -storage.cacheSizeIndexDBTagFilters | size | 0 | The maximum size of the indexdb/tagFiltersToMetricIDs cache. |
| -storage.cacheSizeStorageTSID | size | 0 | The maximum size of the storage/tsid cache. |
| -storage.maxDailySeries | int | 0 | The maximum number of unique series that can be created in 24 hours. Excess series will be discarded and logged. |
| -storage.maxHourlySeries | int | 0 | The maximum number of unique series that can be created in 1 hour. Excess series will be discarded and logged. |
| -storage.maxSeriesPerSecond | int | 2000 | The maximum number of series that can be created per second. Excess series will be discarded and logged. |
| -storage.minFreeDiskSpaceBytes | size | 10000000 | The minimum free disk space for the storage path. If the available space is below this value, the storage stops accepting data. |
| -storage.vminsertConnsShutdownDuration | duration | 25s | The time required for gracefully closing vminsert connections. Controls the shutdown speed to reduce load fluctuations. |
| -storageDataPath | string | "vmstorage-data" | The path to store data. |
| -tls | array | [] | Whether to enable TLS for the specified HTTP listening address. Requires setting the certificate and key files. |
| -tlsCertFile | array | [] | The path to the TLS certificate file. Supports dynamic updates. |
| -tlsCipherSuites | array | [] | The list of TLS cipher suites. |
| -tlsKeyFile | array | [] | The path to the TLS key file. Supports dynamic updates. |
| -tlsMinVersion | array | [] | The minimum TLS version required. Supports TLS10, TLS11, TLS12, TLS13. |
| -version | bool | false | Display the version of VictoriaMetrics. |
| -vminsertAddr | string | ":8400" | The listening address for the vminsert service. |
| -vmselectAddr | string | ":8401" | The listening address for the vmselect service. |