2 minute read

I have been working on my SSB Iceberg Demo for going on 2 months now. This past week I have had a chance to build out some Iceberg Time Travel capabilities. I have the code and samples added to my SSB Iceberg Demo but a more interesting time travel story is within the Fraud Detection Demo I have been running. Let’s take a look at the important details.

First, these are sample queries we can execute in Hue to get snapshot ids, query as of SYSTEM_VERSION, ROLLBACK to a version, and confirm current state of data is from rollback.

-- Describe Table
DESCRIBE FORMATTED fraudulent_txn_iceberg; 

-- Get Current Count
select count(*) from fraudulent_txn_iceberg
 -- 1456146

-- Get Snap Shot Ids
DESCRIBE HISTORY fraudulent_txn_iceberg
-- copy 2 ids,  one older than the other

-- Get Totals Per Card Type As of SnapShot 1 
select card, sum(amount) from fraudulent_txn_iceberg FOR SYSTEM_VERSION AS OF 2163411949573389139 GROUP BY card
  -- mastercard       103930672
  -- americanexpress  105070827
  -- visa             104719497

-- Get Totals Per Card Type As of SnapShot 2
select card, sum(amount) from fraudulent_txn_iceberg FOR SYSTEM_VERSION AS OF 2013237884718568734 GROUP BY card
  -- mastercard       116812083
  -- americanexpress  115538225
  -- visa             116185432
 
-- Get Count as of SnapShot 2  
select count(*) from fraudulent_txn_iceberg FOR SYSTEM_VERSION AS OF 2013237884718568734  
 -- 348732
 
-- Roll back to Snapshot 2
ALTER TABLE fraudulent_txn_iceberg  EXECUTE ROLLBACK(2013237884718568734);

-- Confirm current table Count is Correct
select count(*) from fraudulent_txn_iceberg
 -- 348732
 
-- Show Database Totals match Query Line 15
select card, sum(amount) from fraudulent_txn_iceberg GROUP BY card 
  -- mastercard       116812083
  -- americanexpress  115538225
  -- visa             116185432

Next, lets take a look at some Sql Stream Builder queries using the SnapShot Ids to stream from a point in time or stream between a start and end snapshot:

-- First, get snapshots ids for the iceberg table
/* In hue (hue-impala-iceberg DataWarehouse) execute the following query to get start-snapshot-id report
DESCRIBE HISTORY fraudulent_txn_iceberg; 
*/


-- Next, complete a basic select with snapshot-id
select * from fraudulent_txn_iceberg /*+OPTIONS('snapshot-id'='6619035083895556755')*/;

-- Time travel 1 sec stream starting from snap-shot-id
select * from fraudulent_txn_iceberg /*+OPTIONS('streaming'='true', 'monitor-interval'='1s', 'start-snapshot-id'='4263825941508588099')*/;

-- Select data from start snapshot to end snapshot
select * from fraudulent_txn_iceberg /*+OPTIONS('start-snapshot-id'='4263825941508588099', 'end-snapshot-id'='3724519465921078641')*/;

-- Select data from starting timestamp
select * from fraudulent_txn_iceberg /*+OPTIONS('as-of-timestamp'='1699425703000')*/;
-- convert snapshot history timestamp to unix time + 000 

I think Iceberg is pretty cool and it is definitely here to stay. If you need any more information or help with Cloudera Sql Stream Builder or Iceberg reach out as I am always willing to demo, run a hands on lab, or help you with yours.

Check out Whats New in CDP 7.1.9

Check out Whats new in CSA 1.11.

Check out Streaming SQL Console

CDP Public Cloud SSB

As always, check out the docs:

Cloudera Streaming Analytics DOCS

CDP SSB CSA Flink