Installing Apache Kudu

The Apache Kudu project only publishes source code releases, to deploy Kudu on a cluster follow the steps below to build Kudu from source.

Prerequisites and Requirements

Hardware

  • One or more hosts to run Kudu masters. It is recommended to have either one master (no fault tolerance), or three masters (can tolerate one failure). The number of masters must be odd.

  • One or more hosts to run Kudu tablet servers. When using replication, a minimum of three tablet servers is necessary.

A deployment with an even number of masters provides the same level of fault tolerance as a deployment with one fewer master. For example, both four-master and three-master deployments can only tolerate a single failure; two-master deployments cannot tolerate any failures.
Operating System Requirements
Linux

  • RHEL 6, RHEL 7, RHEL 8, CentOS 6, CentOS 7, CentOS 8, Ubuntu 14.04 (trusty), Ubuntu 16.04 (xenial), Ubuntu 18.04 (bionic), Debian 8 (Jessie), or SLES 12.

  • A kernel and filesystem that support hole punching. Hole punching is the use of the fallocate(2) system call with the FALLOC_FL_PUNCH_HOLE option set. See troubleshooting hole punching for more information.

  • ntp.

  • xfs or ext4 formatted drives.

  • Although not a strict requirement, it’s highly recommended to use nscd to cache both DNS name resolution and static name resolution. See troubleshooting slow DNS lookups for more information.

macOS

  • OS X 10.10 Yosemite, OS X 10.11 El Capitan, macOS Sierra, or macOS High Sierra.

Windows

  • Microsoft Windows is unsupported.

Storage

  • If solid state storage is available, storing Kudu WALs on such high-performance media may significantly improve latency when Kudu is configured for its highest durability levels.

Java

  • JDK 8 is required to build Kudu, but a JRE is not required at runtime except for tests.

Build From Source

Below are the steps for each supported operating system to build Kudu from source.

Known Build Issues

  • It is not possible to build Kudu on Microsoft Windows.

  • A C+11 capable compiler (GCC 4.8) is required.

RHEL or CentOS

RHEL or CentOS 6.6 or later is required to build Kudu from source. To build on a version older than 7.0, the Red Hat Developer Toolset must be installed (in order to have access to a C++11 capable compiler).

  1. Install the prerequisite libraries, if they are not installed.

    $ sudo yum install autoconf automake cyrus-sasl-devel cyrus-sasl-gssapi \
  cyrus-sasl-plain flex gcc gcc-c++ gdb git java-1.8.0-openjdk-devel \
  krb5-server krb5-workstation libtool make openssl-devel patch \
  pkgconfig redhat-lsb-core rsync unzip vim-common which

  2. If building on RHEL or CentOS older than 7.0, install the Red Hat Developer Toolset.

    $ DTLS_RPM=rhscl-devtoolset-3-epel-6-x86_64-1-2.noarch.rpm
$ DTLS_RPM_URL=https://www.softwarecollections.org/repos/rhscl/devtoolset-3/epel-6-x86_64/noarch/${DTLS_RPM}
$ wget ${DTLS_RPM_URL} -O ${DTLS_RPM}
$ sudo yum install -y scl-utils ${DTLS_RPM}
$ sudo yum install -y devtoolset-3-toolchain

  3. Optional: If support for Kudu’s NVM (non-volatile memory) block cache is desired, install the memkind library.

    $ sudo yum install memkind

    If the memkind package provided with the Linux distribution is too old (1.8.0 or newer is required), build and install it from source.

    $ sudo yum install numactl-libs numactl-devel
$ git clone https://github.com/memkind/memkind.git
$ cd memkind
$ ./build.sh --prefix=/usr
$ sudo yum remove memkind
$ sudo make install
$ sudo ldconfig

  4. Optional: Install some additional packages, including ruby, if you plan to build documentation.

    $ sudo yum install doxygen gem graphviz ruby-devel zlib-devel
    If building on RHEL or CentOS older than 7.0, the gem package may need to be replaced with rubygems

  5. Clone the Git repository and change to the new kudu directory.

    $ git clone https://github.com/apache/kudu
$ cd kudu

  6. Build any missing third-party requirements using the build-if-necessary.sh script. Not using the devtoolset will result in Host compiler appears to require libatomic, but cannot find it.

    $ build-support/enable_devtoolset.sh thirdparty/build-if-necessary.sh

  7. Build Kudu, using the utilities installed in the previous step. Choose a build directory for the intermediate output, which can be anywhere in your filesystem except for the kudu directory itself. Notice that the devtoolset must still be specified, else you’ll get cc1plus: error: unrecognized command line option "-std=c++11".

    mkdir -p build/release
cd build/release
../../build-support/enable_devtoolset.sh \
  ../../thirdparty/installed/common/bin/cmake \
  -DCMAKE_BUILD_TYPE=release ../..
make -j4

    If you need to install only a subset of Kudu executables, you can set the following cmake flags to OFF in order to skip any of the executables.

    • KUDU_CLIENT_INSTALL (set to OFF to skip installing /usr/local/bin/kudu executable)

    • KUDU_TSERVER_INSTALL (set to OFF to skip installing /usr/local/sbin/kudu-tserver executable)

    • KUDU_MASTER_INSTALL (set to OFF to skip installing /usr/local/sbin/kudu-master executable)

    E.g., use the following variation of cmake command if you need to install only Kudu client libraries and headers:

    ../../build-support/enable_devtoolset.sh \
  ../../thirdparty/installed/common/bin/cmake \
  -DKUDU_CLIENT_INSTALL=OFF \
  -DKUDU_MASTER_INSTALL=OFF \
  -DKUDU_TSERVER_INSTALL=OFF
  -DCMAKE_BUILD_TYPE=release ../..

  8. Optional: install Kudu executables, libraries and headers.

    Running sudo make install installs the following:

    • kudu-tserver and kudu-master executables in /usr/local/sbin

    • Kudu command line tool in /usr/local/bin

    • Kudu client library in /usr/local/lib64/

    • Kudu client headers in /usr/local/include/kudu

    The default installation directory is /usr/local. You can customize it through the DESTDIR environment variable.

    sudo make DESTDIR=/opt/kudu install

  9. Optional: Build the documentation. NOTE: This command builds local documentation that is not appropriate for uploading to the Kudu website.

    $ make docs
Example 1. RHEL / CentOS Build Script

This script provides an overview of the procedure to build Kudu on a newly-installed RHEL or CentOS host, and can be used as the basis for an automated deployment scenario. It skips the steps marked Optional above.

#!/bin/bash

sudo yum -y install autoconf automake curl cyrus-sasl-devel cyrus-sasl-gssapi \
  cyrus-sasl-plain flex gcc gcc-c++ gdb git java-1.8.0-openjdk-devel \
  krb5-server krb5-workstation libtool make openssl-devel patch pkgconfig \
  redhat-lsb-core rsync unzip vim-common which
DTLS_RPM=rhscl-devtoolset-3-epel-6-x86_64-1-2.noarch.rpm
DTLS_RPM_URL=https://www.softwarecollections.org/repos/rhscl/devtoolset-3/epel-6-x86_64/noarch/${DTLS_RPM}
wget ${DTLS_RPM_URL} -O ${DTLS_RPM}
sudo yum install -y scl-utils ${DTLS_RPM}
sudo yum install -y devtoolset-3-toolchain
git clone https://github.com/apache/kudu
cd kudu
build-support/enable_devtoolset.sh thirdparty/build-if-necessary.sh
mkdir -p build/release
cd build/release
../../build-support/enable_devtoolset.sh \
  ../../thirdparty/installed/common/bin/cmake \
  -DCMAKE_BUILD_TYPE=release \
  ../..
make -j4

Ubuntu or Debian

  1. Install the prerequisite libraries, if they are not installed.

    $ sudo apt-get install autoconf automake curl flex g++ gcc gdb git \
  krb5-admin-server krb5-kdc krb5-user libkrb5-dev libsasl2-dev libsasl2-modules \
  libsasl2-modules-gssapi-mit libssl-dev libtool lsb-release make ntp \
  openjdk-8-jdk openssl patch pkg-config python rsync unzip vim-common

  2. Optional: If support for Kudu’s NVM (non-volatile memory) block cache is desired, install the memkind library.

    $ sudo apt-get install libmemkind0

    If the memkind package provided with the Linux distribution is too old (1.8.0 or newer is required), build and install it from source.

    $ sudo apt-get install libnuma1 libnuma-dev
$ git clone https://github.com/memkind/memkind.git
$ cd memkind
$ ./build.sh --prefix=/usr
$ sudo apt-get remove memkind
$ sudo make install
$ sudo ldconfig

  3. Optional: Install some additional packages, including ruby, if you plan to build documentation.

    $ sudo apt-get install doxygen gem graphviz ruby-dev xsltproc zlib1g-dev

  4. Clone the Git repository and change to the new kudu directory.

    $ git clone https://github.com/apache/kudu
$ cd kudu

  5. Build any missing third-party requirements using the build-if-necessary.sh script.

    $ thirdparty/build-if-necessary.sh

  6. Build Kudu, using the utilities installed in the previous step. Choose a build directory for the intermediate output, which can be anywhere in your filesystem except for the kudu directory itself.

    mkdir -p build/release
cd build/release
../../thirdparty/installed/common/bin/cmake -DCMAKE_BUILD_TYPE=release ../..
make -j4

    If you need to install only a subset of Kudu executables, you can set the following cmake flags to OFF in order to skip any of the executables.

    • KUDU_CLIENT_INSTALL (set to OFF to skip installing /usr/local/bin/kudu executable)

    • KUDU_TSERVER_INSTALL (set to OFF to skip installing /usr/local/sbin/kudu-tserver executable)

    • KUDU_MASTER_INSTALL (set to OFF to skip installing /usr/local/sbin/kudu-master executable)

    E.g., use the following variation of cmake command if you need to install only Kudu client libraries and headers:

      ../../thirdparty/installed/common/bin/cmake \
  -DKUDU_CLIENT_INSTALL=OFF \
  -DKUDU_MASTER_INSTALL=OFF \
  -DKUDU_TSERVER_INSTALL=OFF
  -DCMAKE_BUILD_TYPE=release ../..

  7. Optional: install Kudu executables, libraries and headers.

    Running sudo make install installs the following:

    • kudu-tserver and kudu-master executables in /usr/local/sbin

    • Kudu command line tool in /usr/local/bin

    • Kudu client library in /usr/local/lib64/

    • Kudu client headers in /usr/local/include/kudu

    The default installation directory is /usr/local. You can customize it through the DESTDIR environment variable.

    sudo make DESTDIR=/opt/kudu install

  8. Optional: Build the documentation. NOTE: This command builds local documentation that is not appropriate for uploading to the Kudu website.

    $ make docs
Example 2. Ubuntu / Debian Build Script

This script provides an overview of the procedure to build Kudu on Ubuntu, and can be used as the basis for an automated deployment scenario. It skips the steps marked Optional above.

#!/bin/bash

sudo apt-get -y install autoconf automake curl flex g++ gcc gdb git \
  krb5-admin-server krb5-kdc krb5-user libkrb5-dev libsasl2-dev libsasl2-modules \
  libsasl2-modules-gssapi-mit libssl-dev libtool lsb-release make ntp \
  openjdk-8-jdk openssl patch pkg-config python rsync unzip vim-common
git clone https://github.com/apache/kudu
cd kudu
thirdparty/build-if-necessary.sh
mkdir -p build/release
cd build/release
../../thirdparty/installed/common/bin/cmake \
  -DCMAKE_BUILD_TYPE=release ../..
make -j4

SUSE Linux Enterprise Server

  1. Install the prerequisite libraries, if they are not installed.

    $ sudo zypper install autoconf automake curl cyrus-sasl-devel \
  cyrus-sasl-gssapi flex gcc gcc-c++ gdb git java-1_8_0-openjdk-devel \
  krb5-devel krb5-server libtool lsb-release make ntp openssl-devel patch \
  pkg-config python rsync unzip vim

  2. Optional: If support for Kudu’s NVM (non-volatile memory) block cache is desired, install the memkind library.

    $ sudo zypper install memkind

    If the memkind package provided with the Linux distribution is too old (1.8.0 or newer is required), build and install it from source.

    $ sudo zypper install numactl-libs numactl-devel
$ git clone https://github.com/memkind/memkind.git
$ cd memkind
$ ./build.sh --prefix=/usr
$ sudo zypper remove memkind
$ sudo make install
$ sudo ldconfig

  3. Clone the Git repository and change to the new kudu directory.

    $ git clone https://github.com/apache/kudu
$ cd kudu

  4. Build any missing third-party requirements using the build-if-necessary.sh script.

    $ thirdparty/build-if-necessary.sh

  5. Build Kudu, using the utilities installed in the previous step. Choose a build directory for the intermediate output, which can be anywhere in your filesystem except for the kudu directory itself.

    mkdir -p build/release
cd build/release
../../thirdparty/installed/common/bin/cmake \
  -DCMAKE_BUILD_TYPE=release ../..
make -j4

    If you need to install only a subset of Kudu executables, you can set the following cmake flags to OFF in order to skip any of the executables.

    • KUDU_CLIENT_INSTALL (set to OFF to skip installing /usr/local/bin/kudu executable)

    • KUDU_TSERVER_INSTALL (set to OFF to skip installing /usr/local/sbin/kudu-tserver executable)

    • KUDU_MASTER_INSTALL (set to OFF to skip installing /usr/local/sbin/kudu-master executable)

    E.g., use the following variation of cmake command if you need to install only Kudu client libraries and headers:

      ../../thirdparty/installed/common/bin/cmake \
  -DKUDU_CLIENT_INSTALL=OFF \
  -DKUDU_TSERVER_INSTALL=OFF \
  -DKUDU_MASTER_INSTALL=OFF
  -DCMAKE_BUILD_TYPE=release ../..

  6. Optional: install Kudu executables, libraries and headers.

    Running sudo make install installs the following:

    • kudu-tserver and kudu-master executables in /usr/local/sbin

    • Kudu command line tool in /usr/local/bin

    • Kudu client library in /usr/local/lib64/

    • Kudu client headers in /usr/local/include/kudu

    The default installation directory is /usr/local. You can customize it through the DESTDIR environment variable.

    sudo make DESTDIR=/opt/kudu install
Example 3. SLES Build Script

This script provides an overview of the procedure to build Kudu on SLES, and can be used as the basis for an automated deployment scenario. It skips the steps marked Optional above.

#!/bin/bash

sudo zypper install -y autoconf automake curl cyrus-sasl-devel \
  cyrus-sasl-gssapi flex gcc gcc-c++ gdb git java-1_8_0-openjdk-devel \
  krb5-devel libtool lsb-release make ntp openssl-devel patch \
  pkg-config python rsync unzip vim
git clone https://github.com/apache/kudu
cd kudu
thirdparty/build-if-necessary.sh
mkdir -p build/release
cd build/release
../../thirdparty/installed/common/bin/cmake \
  -DCMAKE_BUILD_TYPE=release \
  ../..
make -j4

macOS

The Xcode package is necessary for compiling Kudu. Some of the instructions below use Homebrew to install dependencies, but manual dependency installation is possible.

After installing Xcode, don’t forget to accept the license and install command-line tools, if it’s not done yet:

$ sudo xcodebuild -license
$ sudo xcode-select --install
macOS Known Issues

Kudu support for macOS is experimental, and should only be used for development. See macOS Limitations & Known Issues for more information.

  1. Install the prerequisite libraries, if they are not installed.

    $ brew install autoconf automake cmake git krb5 libtool openssl pkg-config pstree

  2. Optional: Install some additional packages, including ruby, if you plan to build documentation.

    $ brew install doxygen graphviz ruby
$ brew install gnu-sed --with-default-names #The macOS default sed handles the -i parameter differently

  3. Clone the Git repository and change to the new kudu directory.

    $ git clone https://github.com/apache/kudu
$ cd kudu

  4. Build any missing third-party requirements using the build-if-necessary.sh script.

    $ thirdparty/build-if-necessary.sh

    • If different versions of the dependencies are installed and used when calling thirdparty/build-if-necessary.sh, you may get stuck with output similar to the following:

      ./configure: line 16299: error near unexpected token `newline'
./configure: line 16299: `  PKG_CHECK_MODULES('

      The thirdparty builds may be cached and may reflect the incorrect versions of the dependencies. Ensure that you have the correct dependencies listed in Step 1, clean the workspace, and then try to re-build.

      $ git clean -fdx
$ thirdparty/build-if-necessary.sh

    • Some combinations of Homebrew installations and system upgrades can result with a different kind of error:

      libtool: Version mismatch error.  This is libtool 2.4.6, but the
libtool: definition of this LT_INIT comes from libtool 2.4.2.
libtool: You should recreate aclocal.m4 with macros from libtool 2.4.6
libtool: and run autoconf again.

      As described in this thread, a possible fix is to uninstall and reinstall libtool:

      $ brew uninstall libtool && brew install libtool

  5. Build Kudu. Choose a build directory for the intermediate output, which can be anywhere in your filesystem except for the kudu directory itself.

    mkdir -p build/release
cd build/release
../../thirdparty/installed/common/bin/cmake \
  -DCMAKE_BUILD_TYPE=release \
  -DOPENSSL_ROOT_DIR=/usr/local/opt/openssl \
  ../..
make -j4
Example 4. macOS Build Script

This script provides an overview of the procedure to build Kudu on macOS, and can be used as the basis for an automated deployment scenario. It assumes Xcode and Homebrew are installed.

#!/bin/bash

brew tap homebrew/dupes
brew install autoconf automake cmake git krb5 libtool openssl pkg-config pstree
git clone https://github.com/apache/kudu
cd kudu
thirdparty/build-if-necessary.sh
mkdir -p build/release
cd build/release
../../thirdparty/installed/common/bin/cmake \
  -DCMAKE_BUILD_TYPE=release \
  -DOPENSSL_ROOT_DIR=/usr/local/opt/openssl \
  ../..
make -j4

Installing the C++ Client Libraries

See the Kudu client install section at the bottom of Build From Source above.

Only build against the client libraries and headers (kudu_client.so and client.h). Other libraries and headers are internal to Kudu and have no stability guarantees.

Build the Java Client

Requirements

  • JDK 8

To build the Java client, clone the Kudu Git repository, change to the java directory, and issue the following command:

$ ./gradlew assemble

For more information on building the Java parts of the Kudu project, as well as Eclipse integration, see java/README.md.

Upgrade from a Previous Version of Kudu

Before upgrading, you should read the Release Notes for the version of Kudu that you are about to install. Pay close attention to the incompatibilities, upgrade, and downgrade notes that are documented there.

The following upgrade process is only relevant when you have binaries available.

  1. Prepare the software.

    • Place the new kudu-tserver, kudu-master, and kudu binaries into the appropriate Kudu binary directory.

  2. Upgrade the tablet servers.

    • Set the follower_unavailable_considered_failed_sec configuration to a high value (conservatively, twice the expected restart time) to prevent tablet replicas hosted on restarting tablet servers from being evicted and re-replicated.

      $ ./kudu tserver set_flag <tserver> follower_unavailable_considered_failed_sec 7200

    • Restart one tablet server.

    • Wait for all tablet replicas on the tablet server to finish bootstrapping by viewing /tablets page in the tablet server web UI.

    • Restarting the tablet server will have reset the follower_unavailable_considered_failed_sec configuration. Raise it again as needed.

    • Repeat the previous 3 steps for the remaining tablet servers.

    • Restore the original gflag value of every tablet server (the default is 5 minutes)

      $ ./kudu tserver set_flag <tserver> follower_unavailable_considered_failed_sec 300

      An example for a cluster with three tablet servers A, B, C:

      # Step 1: Set the unavailable time for every tablet server to a large value
$ ./kudu tserver set_flag A follower_unavailable_considered_failed_sec 7200
$ ./kudu tserver set_flag B follower_unavailable_considered_failed_sec 7200
$ ./kudu tserver set_flag C follower_unavailable_considered_failed_sec 7200

# Step 2: Restart the tablet server and reset the gflag one by one
<restart A and wait until A is online>
$ ./kudu tserver set_flag A follower_unavailable_considered_failed_sec 7200
<restart B and wait until B is online>
$ ./kudu tserver set_flag B follower_unavailable_considered_failed_sec 7200
<restart C and wait until C is online>
$ ./kudu tserver set_flag C follower_unavailable_considered_failed_sec 7200

# Step 3: Restore the default gflag value (5 minutes) for every tablet server
$ ./kudu tserver set_flag A follower_unavailable_considered_failed_sec 300
$ ./kudu tserver set_flag B follower_unavailable_considered_failed_sec 300
$ ./kudu tserver set_flag C follower_unavailable_considered_failed_sec 300

  3. Upgrade the master servers.

    • Restart the master server one by one.

Next Steps