You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by "jiraposter@reviews.apache.org (JIRA)" <ji...@apache.org> on 2011/08/05 22:46:32 UTC
[jira] [Commented] (HIVE-2246) Dedupe tables' column schemas from
partitions in the metastore db
[ https://issues.apache.org/jira/browse/HIVE-2246?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13080211#comment-13080211 ]
jiraposter@reviews.apache.org commented on HIVE-2246:
-----------------------------------------------------
bq. On 2011-07-25 06:46:04, Ning Zhang wrote:
bq. > trunk/metastore/scripts/upgrade/mysql/008-HIVE-2246.mysql.sql, line 76
bq. > <https://reviews.apache.org/r/1183/diff/2/?file=26824#file26824line76>
bq. >
bq. > is the CHARSET (latin1) the same as SDS? This will require the user's comments to be in latin1 which prevents UTF chars.
Yes, this charset matches the same ones from the official hive schema for 0.7.0.
bq. On 2011-07-25 06:46:04, Ning Zhang wrote:
bq. > trunk/metastore/scripts/upgrade/mysql/008-HIVE-2246.mysql.sql, line 206
bq. > <https://reviews.apache.org/r/1183/diff/2/?file=26824#file26824line206>
bq. >
bq. > can you also add migration script for derby? we support derby as a default metastore RDBMS as well.
Ok, will do. I will add it in the next-next diff here.
bq. On 2011-07-25 06:46:04, Ning Zhang wrote:
bq. > trunk/metastore/src/java/org/apache/hadoop/hive/metastore/ObjectStore.java, line 1752
bq. > <https://reviews.apache.org/r/1183/diff/2/?file=26825#file26825line1752>
bq. >
bq. > here do you check if the 'alter table' command changes the schema (columns definition)? If it just set a table property, then you don't need to create a new ColumnDescriptor right?
bq. >
bq. > Also if a table's schema got changed, a new CD will be created, but the old partition will still have the old CDs. When we query the old partition, do we use the old partitons's CD or the table's CD?
bq. >
bq. > Also in the above case, when you run 'desc table partition <old_partition>', do you return the old partition's CD or the table's CD?
Good point; I should check whether the table columns have changed; I do this already when altering partitions. I added that in the next diff.
If a table's schema changes, it does not update existing partition CDs. If we ever grab the partition object after the schema change, it will refer to its old CD, not the table's CD. However, when querying tables on the CLI, we almost always use the table's set of columns. E.g., if did:
bq. create table test (a string) partitioned by (p1 string, p2 string);
bq. alter table test add partition(p1=1, p2=1);
bq. # populate the p1=1, p2=1 partition with some data now
bq. alter table test add columns (b string)
bq. select * from test where p1 = 1 and p2 = 1,
it'd use the table's latest schema; i.e., return the column 'a's values and the column 'b' as all NULL.
- Sohan
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/1183/#review1176
-----------------------------------------------------------
On 2011-07-22 05:30:29, Sohan Jain wrote:
bq.
bq. -----------------------------------------------------------
bq. This is an automatically generated e-mail. To reply, visit:
bq. https://reviews.apache.org/r/1183/
bq. -----------------------------------------------------------
bq.
bq. (Updated 2011-07-22 05:30:29)
bq.
bq.
bq. Review request for hive, Ning Zhang and Paul Yang.
bq.
bq.
bq. Summary
bq. -------
bq.
bq. This patch tries to make minimal changes to the API while keeping migration short and somewhat easy to revert.
bq.
bq. The new schema can be described as follows:
bq. - CDS is a table corresponding to Column Descriptor objects. Currently, it only stores a CD_ID.
bq. - COLUMNS_V2 is a table corresponding to MFieldSchema objects, or columns. A Column Descriptor holds a list of columns. COLUMNS_V2 has a foreign key to the CD_ID to which it belongs.
bq. - SDS was modified to reference a Column Descriptor. So SDS now has a foreign key to a CD_ID which describes its columns.
bq.
bq. During migration, we create Column Descriptors for tables in a straightforward manner: their columns are now just wrapped inside a column descriptor. The SDS of partitions use their parent table's column descriptor, since currently a partition and its table share the same list of columns.
bq.
bq. When altering or adding a partition, give it it's parent table's column descriptor IF the columns they describe are the same. Otherwise, create a new column descriptor for its columns.
bq.
bq. When adding or altering a table, create a new column descriptor every time.
bq.
bq. Whenever you drop a storage descriptor (e.g, when dropping tables or partitions), check to see if the related column descriptor has any other references in the table. That is, check to see if any other storage descriptors point to that column descriptor. If none do, then delete that column descriptor. This check is in place so we don't have unreferenced column descriptors and columns hanging around after schema evolution for tables.
bq.
bq.
bq. This addresses bug HIVE-2246.
bq. https://issues.apache.org/jira/browse/HIVE-2246
bq.
bq.
bq. Diffs
bq. -----
bq.
bq. trunk/metastore/scripts/upgrade/mysql/008-HIVE-2246.mysql.sql PRE-CREATION
bq. trunk/metastore/src/java/org/apache/hadoop/hive/metastore/ObjectStore.java 1148945
bq. trunk/metastore/src/model/org/apache/hadoop/hive/metastore/model/MColumnDescriptor.java PRE-CREATION
bq. trunk/metastore/src/model/org/apache/hadoop/hive/metastore/model/MStorageDescriptor.java 1148945
bq. trunk/metastore/src/model/package.jdo 1148945
bq.
bq. Diff: https://reviews.apache.org/r/1183/diff
bq.
bq.
bq. Testing
bq. -------
bq.
bq. Passes facebook's regression testing and all existing test cases. In one instance, before migration, the overhead involved with storage descriptors and columns was ~11 GB. After migration, the overhead was ~1.5 GB.
bq.
bq.
bq. Thanks,
bq.
bq. Sohan
bq.
bq.
> Dedupe tables' column schemas from partitions in the metastore db
> -----------------------------------------------------------------
>
> Key: HIVE-2246
> URL: https://issues.apache.org/jira/browse/HIVE-2246
> Project: Hive
> Issue Type: Improvement
> Components: Metastore
> Reporter: Sohan Jain
> Assignee: Sohan Jain
> Attachments: HIVE-2246.2.patch, HIVE-2246.3.patch
>
>
> Note: this patch proposes a schema change, and is therefore incompatible with the current metastore.
> We can re-organize the JDO models to reduce space usage to keep the metastore scalable for the future. Currently, partitions are the fastest growing objects in the metastore, and the metastore keeps a separate copy of the columns list for each partition. We can normalize the metastore db by decoupling Columns from Storage Descriptors and not storing duplicate lists of the columns for each partition.
> An idea is to create an additional level of indirection with a "Column Descriptor" that has a list of columns. A table has a reference to its latest Column Descriptor (note: a table may have more than one Column Descriptor in the case of schema evolution). Partitions and Indexes can reference the same Column Descriptors as their parent table.
> Currently, the COLUMNS table in the metastore has roughly (number of partitions + number of tables) * (average number of columns pertable) rows. We can reduce this to (number of tables) * (average number of columns per table) rows, while incurring a small cost proportional to the number of tables to store the Column Descriptors.
> Please see the latest review board for additional implementation details.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira