Page MenuHomePhabricator
Create Task
Maniphest T94148

Cassandra secondary indexing problem for large pages
Closed, ResolvedPublic
Actions

Description

When retrieving and storing a revision for a large page, the following Cassandra errors appear (output for Barack_Obama):

ResponseError: Can't index column value of size 1483409 for index null on local_group_default_T_parsoid_html.data
ResponseError: Can't index column value of size 842881 for index null on local_group_default_T_parsoid_dataW4ULtxs1oMqJeY.data

The offending line in Cassandra's code is src/java/org/apache/cassandra/cql3/statements/UpdateStatement.java:135 (ver 2.1.3) and refers to a secondary-index-validation failure. CASSANDRA-3057, CASSANDRA-4240, CASSANDRA-8081 and CASSANDRA-8280 all suggest that data larger than 65K cannot be indexed, but we do not index the value field so we should not be getting this error in the first place.

Another peculiar thing about the error is that it states for index null.

Related Objects

Event Timeline

• mobrovac created this task.Mar 27 2015, 10:37 AM
• mobrovac raised the priority of this task from to High.
• mobrovac updated the task description. (Show Details)
• mobrovac added projects: RESTBase, RESTBase-Cassandra, Services.
• mobrovac added subscribers: • mobrovac, • GWicke, Eevans.
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptMar 27 2015, 10:37 AM
• mobrovac mentioned this in T94056: Explicit oldid loads are much slower than no-oldid loads, even if the oldid specified is the latest oldid.Mar 27 2015, 10:40 AM
• GWicke added a comment.EditedMar 27 2015, 6:43 PM

Fortunately, this does not affect production as the indexes at issue don't exist there.

It is a Cassandra bug, and we'll stop creating those native secondary indexes to avoid this issue for auto-created tables outside of production. As a result, we need to change the way we implement global title / revision listings.

Eevans added a comment.Mar 27 2015, 6:46 PM

Filed: https://issues.apache.org/jira/browse/CASSANDRA-9057

• GWicke added a comment.EditedApr 1 2015, 1:23 AM

Work-arounds / fixes for RB in

• GWicke claimed this task.Apr 1 2015, 1:24 AM
• GWicke set Security to None.
• mobrovac closed this task as Resolved.Apr 21 2015, 12:47 PM

Cassandra's SI's are no longer used in RB, so resolving.

Content licensed under Creative Commons Attribution-ShareAlike (CC BY-SA) 4.0 unless otherwise noted; code licensed under GNU General Public License (GPL) 2.0 or later and other open source licenses. By using this site, you agree to the Terms of Use, Privacy Policy, and Code of Conduct. · Wikimedia Foundation · Privacy Policy · Code of Conduct · Terms of Use · Disclaimer · CC-BY-SA · GPL