phoenix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (PHOENIX-3534) Support multi region SYSTEM.CATALOG table
Date Wed, 11 Jul 2018 15:51:00 GMT

    [ https://issues.apache.org/jira/browse/PHOENIX-3534?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16540273#comment-16540273
] 

ASF GitHub Bot commented on PHOENIX-3534:
-----------------------------------------

Github user twdsilva commented on a diff in the pull request:

    https://github.com/apache/phoenix/pull/303#discussion_r201745438
  
    --- Diff: phoenix-core/src/main/java/org/apache/phoenix/coprocessor/MetaDataEndpointImpl.java
---
    @@ -3642,30 +3596,62 @@ private void dropIndexes(PTable table, Region region, List<ImmutableBytesPtr>
in
                 boolean isCoveredColumn = indexMaintainer.getCoveredColumns().contains(colDropRef);
                 // If index requires this column for its pk, then drop it
                 if (isColumnIndexed) {
    -                // Since we're dropping the index, lock it to ensure
    -                // that a change in index state doesn't
    -                // occur while we're dropping it.
    -                acquireLock(region, indexKey, locks);
                     // Drop the index table. The doDropTable will expand
                     // this to all of the table rows and invalidate the
                     // index table
    -                additionalTableMetaData.add(new Delete(indexKey, clientTimeStamp));
    +                Delete delete = new Delete(indexKey, clientTimeStamp);
                     byte[] linkKey =
                             MetaDataUtil.getParentLinkKey(tenantId, schemaName, tableName,
index
                                     .getTableName().getBytes());
    -                // Drop the link between the data table and the
    +                // Drop the link between the parent table and the
                     // index table
    -                additionalTableMetaData.add(new Delete(linkKey, clientTimeStamp));
    -                doDropTable(indexKey, tenantId, index.getSchemaName().getBytes(), index
    -                        .getTableName().getBytes(), tableName, index.getType(),
    -                    additionalTableMetaData, invalidateList, locks, tableNamesToDelete,
sharedTablesToDelete, false, clientVersion);
    -                invalidateList.add(new ImmutableBytesPtr(indexKey));
    +                Delete linkDelete = new Delete(linkKey, clientTimeStamp);
    +                List<Mutation> tableMetaData = Lists.newArrayListWithExpectedSize(2);
    +                Delete tableDelete = delete;
    +                tableMetaData.add(tableDelete);
    +                tableMetaData.add(linkDelete);
    +                // if the index is not present on the current region make an rpc to drop
it
    --- End diff --
    
    This is to handle dropping indexes on views. We could drop a column on a base table which
is used by an index on view which could be on a separate region.


> Support multi region SYSTEM.CATALOG table
> -----------------------------------------
>
>                 Key: PHOENIX-3534
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-3534
>             Project: Phoenix
>          Issue Type: Bug
>            Reporter: James Taylor
>            Assignee: Thomas D'Silva
>            Priority: Major
>             Fix For: 5.0.0, 4.15.0
>
>         Attachments: PHOENIX-3534.patch
>
>
> Currently Phoenix requires that the SYSTEM.CATALOG table is single region based on the
server-side row locks being held for operations that impact a table and all of it's views.
For example, adding/removing a column from a base table pushes this change to all views.
> As an alternative to making the SYSTEM.CATALOG transactional (PHOENIX-2431), when a new
table is created we can do a lazy cleanup  of any rows that may be left over from a failed
DDL call (kudos to [~lhofhansl] for coming up with this idea). To implement this efficiently,
we'd need to also do PHOENIX-2051 so that we can efficiently find derived views.
> The implementation would rely on an optimistic concurrency model based on checking our
sequence numbers for each table/view before/after updating. Each table/view row would be individually
locked for their change (metadata for a view or table cannot span regions due to our split
policy), with the sequence number being incremented under lock and then returned to the client.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message