Working with Apache Kudu

Introduction

Kudu is a columnar storage manager developed for the Apache Hadoop platform. Kudu shares the common technical properties of Hadoop ecosystem applications: it runs on commodity hardware, is horizontally scalable, and supports highly available operation.

Unlike other storage for big data analytics, Kudu isn't just a file format. It's a live storage system which supports low-latency millisecond-scale access to individual rows. Kudu isn't designed to be an OLTP system, but Fast processing of OLAP workloads.

We are using Kudu to provide the data to our real time dashboards.

Quick Concepts

- Columnar Data Store

- A columnar data store stores data in strongly-typed columns

- you can read a single column, or a portion of that column, while ignoring other columns

- a given column contains only one type of data, pattern-based compression can be orders of magnitude more efficient

- Table has a schema and a totally ordered primary key. A table is split into segments called tablets

- Tablet

- is a contiguous segment of a table, similar to a partition in other data storage engines or relational databases

- replicated on multiple tablet servers

- Tablet Server

- stores and serves tablets to clients

- For a given tablet, one tablet server acts as a leader, and the others act as follower replicas of that tablet

- Only leaders service write requests, while leaders or followers each service read requests

- Leaders are elected using Raft Consensus Algorithm.

- One tablet server can serve multiple tablets, and one tablet can be served by multiple tablet servers

- Master

- The master keeps track of all the tablets, tablet servers, the Catalog Table, and other metadata related to the cluster

- there can only be one acting master (the leader)

- also coordinates metadata operations for clients

- All the master’s data is stored in a tablet, which can be replicated to all the other candidate masters.

- Tablet servers heartbeat to the master at a set interval (the default is once per second).

- Raft Consensus Algorithm

- a means to guarantee fault-tolerance and consistency, both for regular tablets and for master data

- Catalog Table

- central location for metadata of Kudu

- tables, schemas, locations, and states

- tablets, which tablet servers have replicas of each tablet, the tablet’s current state, and start and end keys

- may not be read or written directly

- accessible only via metadata operations exposed in the client API -

- Logical Replication

- Kudu replicates operations, not on-disk data. This is referred to as logical replication, as opposed to physical replication

- inserts and updates do transmit data over the network, deletes do not need to move any data

- compaction, do not need to transmit the data over the network in Kudu, different from storage systems that use HDFS

My Environment:

- Master Nodes

1 Master Node

- Tablet Servers

All 5 Data Nodes as Tablet Servers

- Storage

Different hard drives for the the data and WAL directories of Kudu

--fs_wal_dir and --fs_metadata_dir be placed on a high-performance drives with high bandwidth and low latency

- mirror the drives containing these directories in order to make recovering from a drive failure easier

- --fs_data_dirs be comma-separated list of directories, data will be striped across the directories, let Kudu manage its own striping over multiple devices rather than delegating the striping to a RAID-0 array

- --log_dir

- --block_cache_capacity_mb, Maximum amount of memory allocated to the Kudu Tablet Server’s block cache.

- --memory_limit_hard_bytes, Maximum amount of memory a Tablet Server can consume before it starts rejecting all incoming writes.

- kudu.table.history_max_age_sec, Number of seconds to retain history for tablets in this table.

Setting up Kudu

Kudu source code can be downloaded and build by your own from Apache but i've taken the already built RPM. Kudu is based on master client architecture, you need at least one Master and one or more tablet servers. You can query kudu by using Impala or Presto, we are using Presto.

1- Download Kudu RPM from below, i'm using 1.8

https://github.com/MartinWeindel/kudu-docker-slim/releases

2- Installing configuring Kudu Master

Media Path: /opt/sources/kudu/kudu-1.8.0-1.x86_64.rpm
Master: ha01 x.x.x.210
Web UI: http://x.x.x..210:8051/tablet-servers
RPC Port: x.x.x..210:7051

[admin@ha01]$ sudo rpm -ivh /opt/sources/kudu/kudu-1.8.0-1.x86_64.rpm

installing the package above will create

- kudu user on OS

- log location /var/log/kudu

- configuration location /etc/kudu/conf and /etc/default/kudu-master

3- Configure Master

vim /etc/kudu/conf/master.gflagfile

--log_dir=/opt/datastack/log/kudu

--fs_wal_dir=/opt/datastack/kudu/master

--fs_data_dirs=/opt/datastack/kudu/master

4- Create Related Folders defined in configuration, give ownership/permissions to kudu user and start the services

mkdir -p /opt/datastack/log/kudu

mkdir -p /opt/datastack/kudu/master

chown kudu:kudu -R /opt/datastack/log/kudu

chown kudu:kudu -R /opt/datastack/kudu/master

start the Master services

[admin@ha01]$ sudo systemctl status kudu-master.service
[admin@ha01]$ sudo systemctl enable kudu-master.service
[admin@ha01]$ sudo systemctl start kudu-master.service
[admin@ha01]$ sudo systemctl status kudu-master.service

Verify that Master services are running by using browser

http://x.x.x.210:8051/

5- Check Cluster Health

[admin@ha01 ~]$ su - kudu

-bash-4.2$ kudu cluster ksck x.x.x.210:7051

6- Installing Configuring Kudu Tablet Server

Tablet Servers

dn01 x.x.x.196

dn02 x.x.x.197

dn03 x.x.x.198

dn04 x.x.x.199

dn05 x.x.x.200

WEB UI Tablet : http://x.x.x.198:8050

[admin@dn01]$ sudo rpm -ivh /opt/sources/kudu/kudu-1.8.0-1.x86_64.rpm

7- Configure Tablet Servers

vim /etc/kudu/conf/tserver.gflagfile

sudo mkdir -p /data/datastack/kudu/tserver

sudo mkdir -p /data/datastack/log/kudu

sudo chown kudu:kudu -R /data/datastack/kudu/

sudo chown kudu:kudu -R /data/datastack/log/kudu/

DN01