Open topic with navigation
Where resolution is associated with a reported issue, a Zendesk reference number is displayed, for example (9865).
This section includes fixes that don't change the behavior of the product. For a summary of changes, see Changes in this Release.
This section provides a summary of the issues resolved in the release of NuoDB 3.4.3.
ping-timeoutdatabase option is enabled, and a network partition event occurs when the database is under heavy load, database processes in the majority partition may evict themselves incorrectly.
ping-timeoutdatabase option enabled, if an SM process fails and is restarted before the surviving SM finishes processing the node failure, that process may not be able to join the cluster. If the surviving SM process also fails before the original SM is restarted successfully, the cluster may not be able to be restarted from the capture file. This issue is more likely to be encountered in releases prior to NuoDB 3.4, since the time to process node failures was significantly reduced in NuoDB 3.4. (9815)
INSERToperations on a table with a unique index can cause a Storage Manager to fail. (9865)
SYSTEM.ADMINISTRATORrole, for example a DBA, or the user who created the role can revoke privileges to a role.
ping-timeoutdatabase option enabled and in the case of a network partition resulting in partial connectivity between nodes, it is possible that NuoDB will shut down more database engine processes than necessary to maintain database process quorum. (9935)
INSERT SELECT from dual <clause>ignores any
LIMITclauses on the
This section provides a summary of the issues resolved in the release of NuoDB 3.4.2.
executeBatchcommand with the NuoDB C driver, despite error strings being stored in the corresponding entry of the
INSERT...SELECTqueries on temporary tables (with multiple indexes) with an
INlist containing hundreds of values is slower compared to separate
INSERT...SELECTqueries with a very large
INlist on temporary tables with multiple indexes has been improved.
--initialize argument, the database will not start. (9689)
--initializeargument and the database archive directory is not empty.
IDLE_CONNECTION_TIMEOUTsystem property is configured, the period of time since the last user interaction should be calculated. Some driver-level messages are resetting this period. As a result, connections are not timing out. (9554)
ON COMMIT DELETE ROWS, the TE exits unexpectedly. (9600)
BEFORE UPDATEtriggers, being used to create and use a synthetic table partition column, are unable to change the selected partition for a row causing results to be written to the incorrect partition. (9713)
This section provides a summary of the issues resolved in the release of NuoDB 3.4.1.
nuocmd) can generate an encoding exception if the string has non-ASCII characters and the byte-encoding is not explicitly specified.
Introduced in NuoDB 3.4, NuoDB Admin's nuoadmin Docker entrypoint requires a
NUODB_BOOTSTRAP_SERVERID environment variable to specify a bootstrap server. If not specified, in certain simple peering scenarios, the
NUODB_BOOTSTRAP_SERVERID value falls back to the local server ID, causing a split domain.
This issue has been resolved in that a split domain does not occur in scenarios where
NUODB_DOMAIN_ENTRYPOINT is the address of a specific server (not a load balancer). In scenarios where
NUODB_DOMAIN_ENTRYPOINT is not a specific server, but rather a load balancer,
NUODB_BOOTSTRAP_SERVERID must still be specified to avoid split domains.
In certain situations, a newly started TE may hang when trying to join an existing database.
This issue has been resolved. Additional retry logic has been introduced to prevent a new TE hanging as it attempts to join a database.
SELECT DISTINCTstatements, over multiple columns in a derived table, can return unexpected results.
This section provides a summary of the issues resolved in the release of NuoDB 3.4.
REPLACEcommand can locate and replace only existing fields containing a
NULLvalue if the new value is specified as literal null. If the new value is a complex SQL expression that evaluates to a
REPLACEstatement creates a duplicate row instead.
CASEexpressions result in unexpected errors.
CASEexpressions now return a syntax error.
UNIONclauses does not differentiate between
nuoarchive) incorrectly reports warning messages when schema objects are created using inconsistent case for the schema name.
nuoarchivecommand no longer reports warning messages for the scenario described.
START TRANSACTIONstatement does not start a new transaction if an implicitly started transaction is active and has not made any modifications.
It may not be possible to drop a table if the table is later renamed using a schema with a case different to that used when the table was created. (9441)
SQL> create schema [testschema]; SQL> create table [testschema].test(a int generated always as identity); SQL> rename table [TestSchema].test to [TestSchema].test_save; SQL> drop table test_save; SQL> sequence "testschema.test_save$IDENTITY_SEQUENCE" is referenced by table TestSchema.test_save and cannot be dropped
This issue has been resolved, allowing tables to always be dropped after they have been renamed.
STORE INclause, and the table is altered to add a partition without specifying the
STORE INclause, an error is returned indicating that no default storage group is specified. (8875)
CREATE TABLEcommand included the
STORE INclause, and a partition is added without including the
STORE INclause, the default storage group is used. In addition, a new
DEFAULTSTORAGEGROUPcolumn has been added to PARTITITIONEDTABLES which displays the default storage group for partitioned tables.
getreleaseversion()function collects the release version from every database process before also returning the common release version that all database processes are using. As a result, this function is time consuming. (9428)
getreleaseversion()is executed, is running at the most recent database version, only that release version is returned. In this case,
getreleaseversion()no longer attempts to connect to other database processes.
value) may cause duplicate values to be generated.
GENERATED ALWAYSmay result in the column having duplicate keys after subsequent inserts into the table.
CREATE TABLE, may not always be executed in the same order.
NOT NULLcolumn constraint to an existing table with data succeeds even when the column contains
READ COMMITTEDtransaction isolation and is not rolled back if the user transaction is.
CREATE INDEXstatement with concurrent
DELETEoperations on the same table may lead to duplicate or missing index keys. Or, issuing an
ALTER TABLEstatement with concurrent
DELETEoperations on the same table may lead to values violating check constraints.
ENUMtypes, with different definitions, return invalid results where an index is used for the comparison.
0x01characters returns extra