Alex Behm has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/10116 )

Change subject: IMPALA-6131: Track time of last statistics update in metadata
......................................................................


Patch Set 10:

(5 comments)

http://gerrit.cloudera.org:8080/#/c/10116/10/fe/src/main/java/org/apache/impala/catalog/KuduTable.java
File fe/src/main/java/org/apache/impala/catalog/KuduTable.java:

http://gerrit.cloudera.org:8080/#/c/10116/10/fe/src/main/java/org/apache/impala/catalog/KuduTable.java@244
PS10, Line 244:             Long.toString(System.currentTimeMillis() / 1000));
> It sets it if the property does not exist, but this would not work well for
Makes sense, please add a comment somewhere stating why we prefer to set the 
lastDdlTime in Impala.


http://gerrit.cloudera.org:8080/#/c/10116/10/fe/src/main/java/org/apache/impala/service/CatalogOpExecutor.java
File fe/src/main/java/org/apache/impala/service/CatalogOpExecutor.java:

http://gerrit.cloudera.org:8080/#/c/10116/10/fe/src/main/java/org/apache/impala/service/CatalogOpExecutor.java@785
PS10, Line 785:       msTbl.putToParameters("impala.lastComputeStatsTime",
> I have put the constant to HdfsTable, because all the other property keys r
The Kudu properties are generaly in KuduTable. Which property in HdfsTable also 
applies to Kudu tables?

This new last compute stats time property should be in Table.


http://gerrit.cloudera.org:8080/#/c/10116/10/tests/metadata/test_last_ddl_time_update.py
File tests/metadata/test_last_ddl_time_update.py:

http://gerrit.cloudera.org:8080/#/c/10116/10/tests/metadata/test_last_ddl_time_update.py@74
PS10, Line 74:     def run_test(self, query, expect_changed_ddl_time, 
expect_changed_stats_time):
> The "invalidate metadata %(TBL)s; describe %(TBL)s" combo was used the chec
The vast majority of cases only runs one query at once, and there is no 
fundamental reason to provide a multi-query test interface - all testing can be 
done just as well without it. The multi-query behavior is subtly different than 
running the single-query interface multiple times, so I think overall it's 
simpler to think about a single-query interface.


http://gerrit.cloudera.org:8080/#/c/10116/10/tests/metadata/test_last_ddl_time_update.py@93
PS10, Line 93:       # Hive uses a seconds granularity on the last ddl time.
> Isn't it an HMS convention in this case? Or is there a reason behind not us
I agree it's an HMS convention. Let's change the comment to state that then.


http://gerrit.cloudera.org:8080/#/c/10116/10/tests/metadata/test_last_ddl_time_update.py@155
PS10, Line 155:     h.expect_no_time_change("drop incremental stats %(TBL)s 
partition (j=1, s='2012')")
> Are you sure about this? I use DROP INCREMENTAL STATS on purpose to check +
ok wfm



--
To view, visit http://gerrit.cloudera.org:8080/10116
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I59a671ac29d352bd92ce40d5cb6662bb23f146b5
Gerrit-Change-Number: 10116
Gerrit-PatchSet: 10
Gerrit-Owner: Csaba Ringhofer <[email protected]>
Gerrit-Reviewer: Alex Behm <[email protected]>
Gerrit-Reviewer: Csaba Ringhofer <[email protected]>
Gerrit-Reviewer: Lars Volker <[email protected]>
Gerrit-Reviewer: Zoltan Borok-Nagy <[email protected]>
Gerrit-Comment-Date: Mon, 14 May 2018 20:58:13 +0000
Gerrit-HasComments: Yes

Reply via email to