github.com/1aal/kubeblocks@v0.0.0-20231107070852-e1c03e598921/docs/user_docs/kubeblocks-for-kafka/configuration-recommendations-for-production-environments.md (about)

     1  ---
     2  title: Resource description
     3  description: Kafka resource description
     4  keywords: [kafka, java heap, hardware resource]
     5  sidebar_position: 2
     6  sidebar_label: Resource description
     7  ---
     8  
     9  # Resource Description
    10  
    11  ## Java Heap
    12  
    13  Kafka Server's JVM Heap configuration, production environment can refer to the official recommended configuration:
    14  
    15  ```bash
    16  -Xmx6g -Xms6g -XX:MetaspaceSize=96m -XX:+UseG1GC -XX:MaxGCPauseMillis=20 -XX:InitiatingHeapOccupancyPercent=35 -XX:G1HeapRegionSize=16M -XX:MinMetaspaceFreeRatio=50 -XX:MaxMetaspaceFreeRatio=80 -XX:+ExplicitGCInvokesConcurrent
    17  ```
    18  
    19  - Combined mode
    20      When creating a Kafka Cluster, specify the `--broker-heap` parameter.
    21  - Separated mode
    22      When creating a Kafka Cluster, specify the component parameters with the `--broker-heap`; specify controller with the `--controller-heap` parameter.
    23  
    24  :::note
    25  
    26  When modifying the Java Heap configuration, attention should be paid to the resources allocated to the Cluster at the same time. For example, `--memory=1Gi`, but `-Xms` in `--broker-heap` is specified as `6g`, the broker service will not start normally due to insufficient memory.
    27  
    28  :::
    29  
    30  ## Hardware resources
    31  
    32  It is recommended to use the following hardware resource in a production environment:
    33  
    34  - `-cpu` >= 16 cores
    35  - `-memory` >= 64Gi
    36  
    37  Since Kafka uses a large amount of Page Cache to improve read and write speed, memory resources outside the heap memory allocated can improve overall performance.
    38  
    39  - `-storage` configure it according to the specific situations
    40  
    41  The default compression algorithm for Kafka Cluster is `compression.type=producer`, which is specified by the Producer end. Refer to the average message size of the Topic, compression ratio, number of Topic replicas, and data retention time to perform a comprehensive evaluation.