-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/18626/#review36258
-----------------------------------------------------------



common/src/main/java/org/apache/falcon/catalog/AbstractCatalogService.java
<https://reviews.apache.org/r/18626/#comment67209>

    Why do you need a separate method for this? Why not add this to 
getPartition method itself. Seems wasted IMO.



common/src/main/java/org/apache/falcon/catalog/HiveCatalogService.java
<https://reviews.apache.org/r/18626/#comment67210>

    Why do you need a separate method for this? Why not add this to 
getPartition method itself. Seems wasted IMO.
    
    Also, it naturally belongs to CatalogStorage, no?



retention/src/main/java/org/apache/falcon/retention/FeedEvictor.java
<https://reviews.apache.org/r/18626/#comment67212>

    nit: method can be map or fill rather than get



retention/src/main/java/org/apache/falcon/retention/FeedEvictor.java
<https://reviews.apache.org/r/18626/#comment67213>

    Can this method be on CatalogStorage?



retention/src/main/java/org/apache/falcon/retention/FeedEvictor.java
<https://reviews.apache.org/r/18626/#comment67215>

    nit: dropPartitionsForAnInstance



retention/src/main/java/org/apache/falcon/retention/FeedEvictor.java
<https://reviews.apache.org/r/18626/#comment67216>

    Why 2 booleans? dropped and deleted? rename dropped to deleted.


- Seetharam Venkatesh


On Feb. 28, 2014, 2:56 p.m., Satish Mittal wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/18626/
> -----------------------------------------------------------
> 
> (Updated Feb. 28, 2014, 2:56 p.m.)
> 
> 
> Review request for Falcon and Srikanth Sundarrajan.
> 
> 
> Repository: falcon-git
> 
> 
> Description
> -------
> 
> When an HCatalog based feed is scheduled in falcon, retention only looks at 
> the first partition key that satisfies either of date pattern: yyyy | MM | dd 
> | HH | mm. As a result, it calculates a partition filter that contains only 
> one of these patterns. However if HCatalog table is defined in such a way 
> that date spans across multiple partition keys (year/month/day/hour/minute), 
> then feed retention doesn't delete any partitions that are granular than 
> first level (year).
> 
> 
> Diffs
> -----
> 
>   common/src/main/java/org/apache/falcon/catalog/AbstractCatalogService.java 
> fc9c3b1 
>   common/src/main/java/org/apache/falcon/catalog/HiveCatalogService.java 
> 3c3660e 
>   common/src/main/java/org/apache/falcon/entity/common/FeedDataPath.java 
> 4031e14 
>   retention/src/main/java/org/apache/falcon/retention/FeedEvictor.java 
> 13c447c 
>   
> webapp/src/test/java/org/apache/falcon/lifecycle/TableStorageFeedEvictorIT.java
>  770780e 
> 
> Diff: https://reviews.apache.org/r/18626/diff/
> 
> 
> Testing
> -------
> 
> - Added new integration tests in TableStorageFeedEvictorIT.java to test 
> retention for an Hcatalog feed where date consists of multiple partitions 
> columns (year/month/day).
> - Verified the retention behavior on a test cluster having an Hcatalog based 
> feed partitioned by year/month/day/hour/minute/country.
> 
> 
> Thanks,
> 
> Satish Mittal
> 
>

Reply via email to