Drop table causes ORA-08103 with another one

drop-tableoracleoracle-11gpartitioningtruncate

In my process I have 2 tables.

target table, load table.

The target table holds the current data, and a lot of queries are fired against it.

When new data coming we use a FULL LOAD and the goal is to switch the current data that in the target table with the new data.

The new data exists in the load table.

In goal to do that we are using

Exchange partition

Those are the steps:

create new load table with only 1 partition -all_rows (values less than maxvalue).
insert all the data from source to the load table by insert statement.
exchange partition of the all_rows (=all the new data) with the target table.

That way, we have the most new data, and there is no space that data is not available.
The problem start when I want to drop my load table after the exchange.
If someone send a query to the target table he can get exception:

ORA-08103: Object no longer exists.

Even the drop was only on the load_table. Same when load table truncate.

What does that mean?
Those two tables will be together forever?
Every truncate or drop I need to be afraid on the second one?
Why they are connected at all?

I'm using oracle 11g.

My actions:

CREATE TABLE tbl_ld
partition by range (col1) (partition all_rows values less than (maxvalue)
AS  SELECT * FROM tbl_tgt WHERE 1=2 ;

insert into tbl_ld
select * from tbl_temp;

alter table tbl_ld exchange partition all_rows with table tbl_tgt;

-- drop or truncate
drop table tbl_ld;
truncate table tbl_ld;

Best Answer

AskTom (answered by Connor McDonald) : ORA - 08103:Object No Longer Exists and an even more detailed explanation by Connor McDonald: A cool thing with EXCHANGE PARTITION (recommended by AskTom here).

TRUNCATE TABLE is a DDL statement, not a DML statement like delete. But what happens is a phenomena that is similar to a phenomena that can be caused by delete or update statements.

If delete or update statements are involved then we have

ORA-01555: snapshot too old: rollback segment number string with name "string" too small

There is also an AskTom article for ORA-01555

What happens:

Assume in your transaction you start a SELECT statement on a table. During this select statement another transaction manipulates the data:

Either it deletes or updates the table data (DML) and commits
or it truncates a partitions or the table or exchanges a partition (DDL)

In this case your session that executes the select statement cannot find the data on its original place because the table has already changed. But it is necessary to retrieve the table data in a consistent way, this means not need the data from the table from the point in time when the query started. YOur session it may be able to find the data in a rollback segment (in case of a DML statement) or in the table whose segment you have exchanged with the partition (in case of exchange partition) or it is a segment that is not assigned to a table anymore but is still available and unaltered on disk (DDL statement). If Oracle finds all the data it can continue with your select statement. If it does not find the data, either because the rollback segment where already overwritten (DML case) or the segment was already modified or deleted (DDL case) then Oracle will throw an error. in the DML case Oracle will throw an ORA-01555, in the DDL case it will throw a ORA-08103.

Connor McDonald writes in the first article referenced:

Partitioned tables are the most common scenario to see this error, but by no means the only way.

Anything that physically alters an underlying segment that is being queried can create the issue. Examples:

truncate table

alter table move

alter index rebuild

Both articles are very instructive and I recommend you to read both (the second one is a video)

So how can you handle this problem?

simply repeat the action in case of this error (or rollback the transaction and repeat it)
maybe it helps if you wait some time before you drop the exchanged partition. Maybe this will decrease the frequency of this error.

Related Solutions

Mysql – Recreate partition in thesql

Running TRUNCATE PARTITION works fine

DROP DATABASE IF EXISTS partest;
CREATE DATABASE partest;
USE partest
CREATE TABLE rolando (
  id int not null auto_increment,
  LogDate DATE,
  PRIMARY KEY (id,LogDate)
) ENGINE=InnoDB DEFAULT CHARSET=utf8
PARTITION BY RANGE (DAYOFYEAR(`LogDate`))
(PARTITION p1 VALUES LESS THAN (8) ENGINE = InnoDB,
 PARTITION p2 VALUES LESS THAN (15) ENGINE = InnoDB,
 PARTITION p3 VALUES LESS THAN (22) ENGINE = InnoDB,
 PARTITION p4 VALUES LESS THAN (29) ENGINE = InnoDB,
 PARTITION p5 VALUES LESS THAN (36) ENGINE = InnoDB,
 PARTITION p6 VALUES LESS THAN (43) ENGINE = InnoDB,
 PARTITION p7 VALUES LESS THAN (50) ENGINE = InnoDB,
 PARTITION p8 VALUES LESS THAN (57) ENGINE = InnoDB,
 PARTITION p9 VALUES LESS THAN (64) ENGINE = InnoDB,
 PARTITION p51 VALUES LESS THAN (358) ENGINE = InnoDB,
 PARTITION p52 VALUES LESS THAN MAXVALUE ENGINE = InnoDB);
ALTER TABLE rolando TRUNCATE PARTITION p4;

Here is the output

mysql> DROP DATABASE IF EXISTS partest;
Query OK, 1 row affected (1.55 sec)

mysql> CREATE DATABASE partest;
Query OK, 1 row affected (0.00 sec)

mysql> USE partest
Database changed
mysql> CREATE TABLE rolando (
    ->   id int not null auto_increment,
    ->   LogDate DATE,
    ->   PRIMARY KEY (id,LogDate)
    -> ) ENGINE=InnoDB DEFAULT CHARSET=utf8
    -> PARTITION BY RANGE (DAYOFYEAR(`LogDate`))
    -> (PARTITION p1 VALUES LESS THAN (8) ENGINE = InnoDB,
    ->  PARTITION p2 VALUES LESS THAN (15) ENGINE = InnoDB,
    ->  PARTITION p3 VALUES LESS THAN (22) ENGINE = InnoDB,
    ->  PARTITION p4 VALUES LESS THAN (29) ENGINE = InnoDB,
    ->  PARTITION p5 VALUES LESS THAN (36) ENGINE = InnoDB,
    ->  PARTITION p6 VALUES LESS THAN (43) ENGINE = InnoDB,
    ->  PARTITION p7 VALUES LESS THAN (50) ENGINE = InnoDB,
    ->  PARTITION p8 VALUES LESS THAN (57) ENGINE = InnoDB,
    ->  PARTITION p9 VALUES LESS THAN (64) ENGINE = InnoDB,
    ->  PARTITION p51 VALUES LESS THAN (358) ENGINE = InnoDB,
    ->  PARTITION p52 VALUES LESS THAN MAXVALUE ENGINE = InnoDB);
Query OK, 0 rows affected (2.60 sec)

mysql> ALTER TABLE rolando TRUNCATE PARTITION p4;
Query OK, 0 rows affected (0.25 sec)

mysql>

OK, fine. It works. However, doing this is logically unsafe. Why ? The partitioned table is based on DAYOFYEAR(datetime_col)). Since each partition contains a week for multiple years, running alter table t1 truncate partition p4; would wipe out week 4 from multiple years.

OK, let's try dropping and adding the partition

mysql> ALTER TABLE rolando DROP PARTITION p4;
Query OK, 0 rows affected (0.41 sec)
Records: 0  Duplicates: 0  Warnings: 0

mysql> ALTER TABLE rolando ADD PARTITION
    -> (PARTITION p4 VALUES LESS THAN (29) ENGINE = InnoDB);
ERROR 1481 (HY000): MAXVALUE can only be used in last partition definition
mysql>

This makes sense since MAXVALUE already exists in the definition. This prevents you from adding (appending) a partition if MAXVALUE is present.

While you could play games with ALTER TABLE ... REORGANIZE PARTITION, the complexity isn't worth it and you won't achieve the result of wedging in a partition.

Rather that explanining further partition machinations, just do the following

Make sure an index exists on the date field
Perform the DELETE SQL already mentioned

An alternative would be to create a table with a weekly partition then drop the exact week.

Mysql – How to efficiently partition a large MySQL (Aurora) table with multiple search columns

I would consider using a date range as the partitioning scheme, and splitting by week which you can do using a function against the timestamp field. Using days would create too many partitions, and months would not help your queries much especially when the range spans two months.

Using the range partition, you can create mutliple partitions covering your active time period, and out into the future for a year, then nearer the end of that range you can add partitions to the schema as required.
If necessary, you can also drop the old partitions when they are no longer needed.

I can't prepare a sample partition schema for you, as I'm running short of time at the moment, but the MySQL docs have pretty good coverage for this type of partitioning.

Hope that helps,
Dave

Best Answer

Related Solutions

Mysql – Recreate partition in thesql

Mysql – How to efficiently partition a large MySQL (Aurora) table with multiple search columns

Related Question