| A parse error may occur after a request receives an error and the request is restarted. |
Replication |
|
|
MXSHR-5247 |
5.8.00 |
|
Downstream systems that consume Kafka use sv_tran_id to dictate order of processing. This may cause changes to be processed out of order if multiple transactions interleave. For example, say there are 2 sessions A,B:
A: begin transaction
B: begin transaction (higher sv_tran_id
B: commit
A: commit and overrides B's changes
The changes for B will come after A and processed changes out of order.
The solution is to include an addition field (sv_commit_tran_id) which is the SCN from Oracle when commit occurred. Changes can then be sorted by sv_tran_id, sv_commit_tran_id.
|
Director, Preparer |
|
|
MXSHR-5194 |
5.8.00 |
| Duplicate transactions may be saved in the metabase when replicating in memory and a target apply error occurs. |
Replication |
|
|
MXSHR-5277 |
5.8.00 |
| For Oracle RAC, RP_RESTART may not be updated correctly. When the kernel restarts duplicate transactions may be written to RP_LOG. |
Replication from Oracle RAC servers |
|
|
MXSHR-5298 |
5.8.00 |
| IBMi Change Selector terminate with error while processing one journal entry that contains invalid value (here binary zeros) for Zoned Decimal column. |
Change Selector |
A customer does a programmatic update of a table record pushing some invalid value into a Zoned Decimal column. In consequence Change Selector fails to parse the value and terminate with an error. |
Improved parsing method for Zoned Decimal that will handle invalid values |
MXSHR-5343 |
5.8.00 |
| If the metabase is installed in IASP then the kernel failed to start. |
Kernel start |
|
Added the IASP context if it is available so that metabase can be fetched appropriately |
MXSHR-5306 |
5.8.00 |
| Messages in RP_MSG in IBMi metabase could have the same timestamp and be processed in the wrong order by the Change Selector. Occurs particularly while doing a model update in MonCon sending deregister/register table messages within the same microsecond. |
IBMi Change Selector |
Since a table can be registered and then deregistered (wrong order) the table is finally not processed and data not replicated. |
Combinate timestamp with a sequential unique ID for each message and ORDER BY when retrieving |
MXSHR-4870 |
5.8.00 |
| Oracle returns a deadlock error when dropping or creating a log group during enable/disable capture. |
Enable/Disable capture for a table or request |
|
|
MXSHR-4852 |
5.8.00 |
| Orphan commit rows may occur if a transaction spans multiple requests and all rows for a request are backed out. |
Replication |
|
|
MXSHR-5258 |
5.8.00 |
| slowness for copy request to Azure target |
kernel |
table constraints not dropped prior to loading/inserting data to target table |
remove constraints prior to load/insert of data |
MXSHR-5203 |
5.8.00 |
| The algorithm to calculate length of number with precision when using TeraData MLOAD utility is not correct. The algorithm needs to account for implied zero left of decimal when value is less than zero. |
Copy Applier |
|
|
MXSHR-5513 |
5.8.00 |
| The replication request randomly aborts without meaningful error message when replicating to Snowflake target. |
Replication to Snowflake target |
|
|
MXSHR-5541 |
5.8.00 |
| Transactions may be sent out of order subsequent transactions with earlier commit timestamps may be skipped. |
Replication for Oracle RAC servers |
|
|
MXSHR-5215 |
5.8.00 |
| When replicating from Oracle RAC transaction processed from the temp table may not set the commit timestamp correctly causing skipped transactions. |
Replication from Oracle RAC servers |
|
|
MXSHR-5283 |
5.8.00 |