Weird memory usage compared to JDBC and BigQuery kafka connects

[nicolae-gorgias]

Describe the bug

Hey team, we are using ClickHouse, JDBC and BigQuery Kafka Connects with identical settings and resources:

• 4GB RAM limit and requests in k8s
• 2CPU limit and requests
• KAFKA_HEAP_OPTS : -XX:+UseContainerSupport -XX:InitialRAMPercentage=60 -XX:MaxRAMPercentage=80

I found that only ClickHouse has a weird memory usage reported on max:kubernetes.memory.usage_pct metric by growing until 100%, you can see it here, red is CH KC (90%-100%), blue is BQ KC (80%)

Memory pool also looks unstable compared to BigQuery KC (violet - CH, blue - BQ):

This command returns same settings on both containers:

java -XX:+UseContainerSupport -XX:InitialRAMPercentage=60 -XX:MaxRAMPercentage=80 -XshowSettings:VM -version
VM settings:
    Max. Heap Size (Estimated): 3.20G
    Using VM: OpenJDK 64-Bit Server VM

Property settings:
    awt.toolkit = sun.awt.X11.XToolkit
    file.encoding = UTF-8
    file.separator = /
    java.awt.graphicsenv = sun.awt.X11GraphicsEnvironment
    java.awt.printerjob = sun.print.PSPrinterJob
    java.class.path = 
    java.class.version = 55.0
    java.home = /usr/lib/jvm/zulu11-ca
    java.io.tmpdir = /tmp
    java.library.path = /usr/java/packages/lib
        /usr/lib64
        /lib64
        /lib
        /usr/lib
    java.runtime.name = OpenJDK Runtime Environment
    java.runtime.version = 11.0.18+10-LTS
    java.specification.name = Java Platform API Specification
    java.specification.vendor = Oracle Corporation
    java.specification.version = 11
    java.vendor = Azul Systems, Inc.
    java.vendor.url = http://www.azul.com/
    java.vendor.url.bug = http://www.azul.com/support/
    java.vendor.version = Zulu11.62+17-CA
    java.version = 11.0.18
    java.version.date = 2023-01-17
    java.vm.compressedOopsMode = Zero based
    java.vm.info = mixed mode
    java.vm.name = OpenJDK 64-Bit Server VM
    java.vm.specification.name = Java Virtual Machine Specification
    java.vm.specification.vendor = Oracle Corporation
    java.vm.specification.version = 11
    java.vm.vendor = Azul Systems, Inc.
    java.vm.version = 11.0.18+10-LTS
    jdk.debug = release
    jdk.vendor.version = Zulu11.62+17-CA
    line.separator = \n 
    os.arch = amd64
    os.name = Linux
    os.version = 5.15.107+
    path.separator = :
    sun.arch.data.model = 64
    sun.boot.library.path = /usr/lib/jvm/zulu11-ca/lib
    sun.cpu.endian = little
    sun.cpu.isalist = 
    sun.io.unicode.encoding = UnicodeLittle
    sun.java.launcher = SUN_STANDARD
    sun.jnu.encoding = UTF-8
    sun.management.compiler = HotSpot 64-Bit Tiered Compilers
    sun.os.patch.level = unknown
    user.dir = /home/appuser
    user.home = /home/appuser
    user.language = en
    user.name = appuser
    user.timezone = 

Operating System Metrics:
    Provider: cgroupv2
    Effective CPU Count: 2
    CPU Period: 100000us
    CPU Quota: 200000us
    CPU Shares: 2048us
    List of Processors: N/A
    List of Effective Processors, 8 total: 
    0 1 2 3 4 5 6 7 
    List of Memory Nodes: N/A
    List of Available Memory Nodes, 1 total: 
    0 
    Memory Limit: 4.00G
    Memory Soft Limit: 0.00K
    Memory & Swap Limit: 4.00G
    Maximum Processes Limit: 629145

openjdk version "11.0.18" 2023-01-17 LTS
OpenJDK Runtime Environment Zulu11.62+17-CA (build 11.0.18+10-LTS)
OpenJDK 64-Bit Server VM Zulu11.62+17-CA (build 11.0.18+10-LTS, mixed mode)

Expected behaviour

ClickHouse Kafka Connect should respect MaxRAMPercentage and stop growing memory usage up to 80%. Do we override some memory settings or have custom ones in ClickHouse Kafka Connect to explain this behavior?

Configuration

Environment

• Kafka-Connect version: 0.0.14
• OS: linux

[Paultagoras]

@nicolae-gorgias I don't believe we have specific memory settings, but one thing I would suggest is updating the version - we've just released 1.0.8 and (especially compared to 0.0.14) there have been a lot of dependencies updated since then 😄

[mshustov]

Echoing what @Paultagoras said: the last version uses the latest JDBC driver version - v0.5.0, which had been refactored to reduce the memory footprint. @nicolae-gorgias would you mind upgrading and sharing your feedback? If you see the problem still exists, we would happily investigate.

[nicolae-gorgias]

@Paultagoras @mshustov thanks, I'll do it and come back with feedback!

[nicolae-gorgias]

@mshustov @Paultagoras Interesting, the vertical line is the upgrade moment, it changed the memory pool metric but total memory is still the same. What other metrics can I check to understand which components consume it?

[Paultagoras]

Hmm interesting - I'll have to take another look. We don't control/configure memory usage to my knowledge, that's all handled by Kafka Connect, but maybe there's some flag or something that we need to provide somewhere.

Is total memory still hitting 100%?