Super Lightweight Design for Real-time Ingestion
The magic that comes with PubSub BigQuery Subscription
What is TTL (Time-to-live)
| name | address | expires_at |
|---|---|---|
| Dummy Dummy | Abc Mountain Lane | 2020-10-29T08:00:00Z |
| Dummy Dummy2 | Def Mountain Lane | 2020-10-29T09:00:00Z |
Use Cases
Manage Sensitive Data
| name | address | expires_at |
|---|---|---|
Only keep most recent data
| id | name | address |
|---|---|---|
| 1 | Dummy Dummy | Abc Mountain Lane |
| 2 | Dummy Dummy2 | Def Mountain Lane |
| id | datetime | amount | expires_at | customer_id |
|---|---|---|---|---|
| 3 | 2020-10-29T08:00:00Z | 13.00 | 2020-11-29T08:00:00Z | 2 |
Customer
Transaction
Sync with another data source
| id | name | address |
|---|---|---|
| 1 | Dummy Dummy | Abc Mountain Lane |
| 2 | Dummy Dummy2 | Def Mountain Lane |
Customer
| id | name | address | expires_at |
|---|---|---|---|
| 1 | Dummy Dummy | Abc Mountain Lane | [datetime + 2 days] |
| 2 | Dummy Dummy2 | Def Mountain Lane | [datetime + 2 days] |
Customer
Daily sync
Let's see how it works
Summary of Caveats
It is difficult to check if the TTL policy has been successfully set because there is no validation
It's taking a long time after expiration timestamp for records to be cleaned up (typically within 72 hours)
Unclear how the policy affects multiple namespaces
Setup a TTL policy takes quite a long time (15 minutes +)
Super Lightweight Design for Real-time Ingestion
By Richard He
Super Lightweight Design for Real-time Ingestion
The magic comes with PubSub BigQuery Subscription
