Attention: Deprecation notice for Bintray, JCenter, GoCenter and ChartCenter. Learn More

milvus/milvus

Chart version: 0.10.6

Api version: v1

App version: 0.10.6

application

Chart Type

Active

Status

Unknown

License

2349

Downloads

Set me up:

Install Chart:

Versions (0)

ReadMe

Chart Info

Used By

Dependencies

Security

Milvus Helm Chart

For more information about installing and using Helm, see the Helm Docs. For a quick introduction to Charts, see the Chart Guide.

To install Milvus, refer to Milvus installation.

Introduction

This chart bootstraps Milvus deployment on a Kubernetes cluster using the Helm package manager.

Prerequisites

Kubernetes 1.10+
Helm >= 2.12.0

Installing the Chart

Add the stable repository bash $ helm repo add stable https://charts.helm.sh/stable $ helm repo add milvus https://milvus-io.github.io/milvus-helm/ 2. Update charts repositories “`

$ helm repo update


3. Install Helm package

To install the chart with the release name `my-release`:

```bash
# Helm v2.x
$ helm install --name my-release milvus/milvus

# Helm v3.x
$ helm install my-release milvus/milvus

Tip: To list all releases, using helm list.

Deploying Milvus with cluster enabled

$ helm install --set cluster.enabled=true --set persistence.enabled=true my-release  .

NOTE: Since all Pods should have the same collection of Milvus files, it is recommended to create just one PV that is shared. This is controlled by setting persistence.enabled=true. You will have to ensure yourself the PVC are shared properly between your pods: - If you are on AWS, you can use Elastic File System (EFS). - If you are on Azure, you can use Azure File Storage (AFS).

To share a PV with multiple Pods, the PV needs to have accessMode ‘ReadOnlyMany’ or ‘ReadWriteMany’.

Uninstall the Chart

To uninstall/delete the my-release deployment:

# Helm v2.x
$ helm delete my-release

# Helm v3.x
$ helm uninstall my-release

The command removes all the Kubernetes components associated with the chart and deletes the release.

Configuration

Milvus server Configuration

The following table lists the configurable parameters of the Milvus chart and their default values.

Parameter	Description	Default
`version`	Configuration Version	`0.5`
`primaryPath`	Primary directory used to save meta data, vector data and index data.	`/var/lib/milvus`
`timeZone`	Use UTC-x or UTC+x to specify a time zone.	`UTC+8`
`autoFlushInterval`	The interval, in seconds, at which Milvus automatically flushes data to disk. 0 means disable the regular flush. (s)	`1`
`fileCleanupTimeout`	The time gap between marking a file as ‘deleted’ and physically deleting this file from disk, range [0, 3600]. (s)	`10`
`logs.path`	Absolute path to the folder holding the log files.	`/var/lib/milvus/logs`
`logs.maxLogFileSize`	The maximum size of each log file, size range [512, 4096]. (MB)	`1024MB`
`logs.logRotateNum`	The maximum number of log files that Milvus keeps for each logging level, num range [0, 1024], 0 means unlimited.	`0`
`cache.insertBufferSize`	Maximum insert buffer size allowed (GB)	`1GB`
`cache.cacheSize`	Size of CPU memory used for cache (GB)	`4GB`
`network.httpPort`	Port that Milvus web server monitors.	`19121`
`wal.enabled`	Enable write-ahead logging.	`true`
`wal.recoveryErrorIgnore`	Whether to ignore logs with errors that happens during WAL	`true`
`wal.bufferSize`	Sum total of the read buffer and the write buffer. (MB)	`256MB`
`wal.path`	Location of WAL log files.	`/var/lib/milvus/db/wal`
`gpu.enabled`	Enable GPU resources	`false`
`gpu.cacheSize`	Size of GPU memory per card used for cache (GB)	`1GB`
`gpu.gpuSearchThreshold`	GPU search threshold	`1000`
`gpu.searchDevices`	Define the GPU devices used for search computation	`[gpu0]`
`gpu.buildIndexDevices`	Define the GPU devices used for index building	`[gpu0]`
`metrics.enabled`	Set this to `true` to enable exporting Prometheus monitoring metrics	`false`
`metrics.address`	Pushgateway address	`127.0.0.1`
`metrics.port`	Prometheus monitoring metrics port	`9091`
`readonly.logs.path`	Absolute path to the folder holding the log files.	`/var/lib/milvus/logs`
`readonly.logs.maxLogFileSize`	The maximum size of each log file, size range [512, 4096]. (MB)	`1024`
`readonly.logs.logRotateNum`	The maximum number of log files that Milvus keeps for each logging level, num range [0, 1024], 0 means unlimited.	`0`
`readonly.cache.insertBufferSize`	Maximum insert buffer size allowed (GB)	`1GB`
`readonly.cache.cacheSize`	Size of CPU memory used for cache (GB)	`4GB`
`readonly.gpu.enabled`	Enable GPU resources	`false`
`readonly.gpu.cacheSize`	Size of GPU memory per card used for cache (GB)	`1GB`
`readonly.gpu.gpuSearchThreshold`	GPU search threshold	`1000`
`readonly.gpu.searchDevices`	Define the GPU devices used for search computation	`[gpu0]`
`readonly.gpu.buildIndexDevices`	Define the GPU devices used for index building	`[gpu0]`
`mishards.debug`	Choose if to enable Debug work mode.	`true`
`mishards.discoveryClassName`	Under the plug-in search path, search the class based on the class name, and instantiate it. Currently, the system provides 2 classes: static and kubernetes.	`kubernetes`
`mishards.trace.enabled`	Enable Mishards tracing service.	`false`
`mishards.trace.tracerClassName`	Under the plug-in search path, search the class based on the class name, and instantiate it. Currently, only Jaeger is supported.	`Jaeger`
`mishards.trace.tracingReportingHost`	The host of the tracing service.	`jaeger`
`mishards.trace.tracingReportingPort`	The port of the tracing service.	`5775`

Milvus Deployment Configuration