Using Distributed Timestamp Global Sequence in Dble: Environment Setup and Verification

Dble provides four global sequence modes; this guide focuses on the distributed‑timestamp method, which generates a globally unique 63‑bit binary ID using Zookeeper as a distributed ID generator.

The 63‑bit ID consists of fields a‑e: a (9‑bit thread ID), b (5‑bit instance ID from INSTANCEID ), c (4‑bit data‑center ID from CLUSTERID ), d (6‑bit auto‑increment), and e (39‑bit low part of the current timestamp).

Environment setup includes deploying a Zookeeper cluster that manages three Dble nodes (dble‑1, dble‑2, dble‑3) and configuring each node’s sequence_distributed_conf.properties :

INSTANCEID=zk
CLUSTERID=01   // for dble‑1
START_TIME=2010-11-04 09:42:54

INSTANCEID=zk
CLUSTERID=02   // for dble‑2
START_TIME=2010-11-04 09:42:54

INSTANCEID=zk
CLUSTERID=03   // for dble‑3
START_TIME=2010-11-04 09:42:54

Update schema.xml to define a table using the distributed sequence and set sequenceHandlerType to 3 in server.xml :

<schema name="schema1" dataNode="dn1">
    <table name="tb_autoIncre" dataNode="dn1,dn2,dn3,dn4" rule="hash-four" cacheKey="id" incrementColumn="id"/>
</schema>

<system>
    <property name="sequenceHandlerType">3</property>
</system>

After editing, execute reload @@config_all on the management port and restart all three Dble instances so the configuration takes effect.

Operation and verification :

mysql -p111111 -utest -P8066 -Dschema1 -e "create table tb_autoIncre(id bigint,time char(120));"

datestr=`date +%Y%m%d`
mysql -p111111 -utest -P8066 -Dschema1 -e "insert tb_autoIncre values('${datestr}');"

mysql -p111111 -utest -P8066 -Dschema1 -e "select * from tb_autoIncre;"

Retrieve the generated id , convert it to binary with select conv(id,10,2) , and dissect the bits to confirm that the cluster ID, instance ID, and timestamp portions match the configured values. Converting the 39‑bit timestamp back to a date (using set @unixtime=.../1000; select from_unixtime(@unixtime); ) yields a time close to the original START_TIME plus the calculated offset, proving the ID generation is correct.

Conclusion: the distributed‑timestamp global sequence in Dble reliably produces positive, globally unique IDs whose components correspond to the configured instance, cluster, and timestamp fields.