[
https://issues.apache.org/jira/browse/IGNITE-24212?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Vladimir Steshin updated IGNITE-24212:
--------------------------------------
Description:
*Consider:*
1) There is a working non-persistent cluster with predefined caches in
_IgniteConfiguration_.
2) The caches have no SQL-entities set.
3) A node creates a table over one of those cache with `_CREATE TABLE IF NOT
EXISTS_`. A SQL table and a schema appear in the cluster.
4) One of the nodes restarts.
5) Next `_SELECT_` queries to the table may fail with '_TABLE NOT FOUND'_
6) Restarted node has no tables in system view '_TABLES_'.
*Cause:*
The node restarts with predefined cache configuration having no SQL entities
set. When node joins cluster, dynamic tables/schema seem not merged or
validated. The restarting node takes the predefined cache configuration as is,
without receiving the dynamically created schema and table. So, the node has no
tables whereas the other cluster does.
+This works+ with the persistence because node stores and reads the cache meta
holding the schema/entities.
*Workarounds:*
1) Do not predefine caches, create them with `_CREATE TABLE_`
2) Set matching SQL entities in the predefined cache config.
3) Use other cache names with `_CREATE TABLE_` (_WITH "CACHE_NAME=._..).
*Suggestions:*
1) Validate schema/caches and do not allow node join cluster.
or
2) Merge static and dynamic schemas/cache configs and send them to the joining
node.
was:
*Consider:*
1) There is a working non-persistent cluster with predefined caches in
_IgniteConfiguration_.
2) The caches have no SQL-entities set.
3) A node creates a table over one of those cache with `_CREATE TABLE IF NOT
EXISTS_`. A SQL table and a schema appear in the cluster.
4) One of the nodes restarts.
5) Next `_SELECT_` queries to the table may fail with '_TABLE NOT FOUND'_
6) Restarted node has no tables in system view '_TABLES_'.
*Cause:*
The node restarts with predefined cache configuration having no SQL entities
set. When node joins cluster, dynamic tables/schema seem not merged or
validated. The restarting node takes the predefined cache configuration as is,
without receiving the dynamically created schema and table. So, the +node has
no tables whereas the othe cluster does+.
+This works+ with the persistence because node stores and reads the cache meta
holding the schema/entities.
*Workarounds:*
1) Do not predefine caches, create them with `_CREATE TABLE_`
2) Set matching SQL entities in the predefined cache config.
3) Use other cache names with `_CREATE TABLE_` (_WITH "CACHE_NAME=._..).
*Suggestions:*
1) Validate schema/caches and do not allow node join cluster.
or
2) Merge static and dynamic schemas/cache configs and send them to the joining
node.
> Non-persistent SQL schema isn't merged or validated when node joins cluster.
> ----------------------------------------------------------------------------
>
> Key: IGNITE-24212
> URL: https://issues.apache.org/jira/browse/IGNITE-24212
> Project: Ignite
> Issue Type: Bug
> Affects Versions: 2.14, 2.16
> Reporter: Vladimir Steshin
> Priority: Minor
> Labels: ise, sql
> Attachments: TestNonPersistentNodeRestartsWithDynamicSQLTable.java
>
>
> *Consider:*
> 1) There is a working non-persistent cluster with predefined caches in
> _IgniteConfiguration_.
> 2) The caches have no SQL-entities set.
> 3) A node creates a table over one of those cache with `_CREATE TABLE IF NOT
> EXISTS_`. A SQL table and a schema appear in the cluster.
> 4) One of the nodes restarts.
> 5) Next `_SELECT_` queries to the table may fail with '_TABLE NOT FOUND'_
> 6) Restarted node has no tables in system view '_TABLES_'.
> *Cause:*
> The node restarts with predefined cache configuration having no SQL entities
> set. When node joins cluster, dynamic tables/schema seem not merged or
> validated. The restarting node takes the predefined cache configuration as
> is, without receiving the dynamically created schema and table. So, the node
> has no tables whereas the other cluster does.
> +This works+ with the persistence because node stores and reads the cache
> meta holding the schema/entities.
> *Workarounds:*
> 1) Do not predefine caches, create them with `_CREATE TABLE_`
> 2) Set matching SQL entities in the predefined cache config.
> 3) Use other cache names with `_CREATE TABLE_` (_WITH "CACHE_NAME=._..).
> *Suggestions:*
> 1) Validate schema/caches and do not allow node join cluster.
> or
> 2) Merge static and dynamic schemas/cache configs and send them to the
> joining node.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
