Konstantin Orlov created IGNITE-18878:
-----------------------------------------
Summary: Sql. Introduce schema synchronisation for query processing
Key: IGNITE-18878
URL: https://issues.apache.org/jira/browse/IGNITE-18878
Project: Ignite
Issue Type: Improvement
Components: sql
Reporter: Konstantin Orlov
h3. Problem
Currently, different nodes may see different version of sql schema. This causes
errors during query execution, because engine can't find required objects (like
indexes) in the schema.
h3. Proposed solution
The proper solution to this will be implemented as part of IGNITE-18733, but
for now let's introduce a quick fix to work this problem around.
The general idea is to have monotonically increasing version of the SQL schema.
This version should be saved at the beginning of the query, and then propagated
to the all node as part of QueryStartRequest. On remote node the query have to
wait until current version at least the same as expected.
h3. Implementation notes
* Need to get sure current schema object has copy-on-write semantic
* MetaStorage revision could be used as a version of the schema. Seems it has
all necessary guarantee
* Previous solution for table seems can be removed (take a look at usage of
{{SqlSchemaManager#tableById}})
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
