Re: commit or not commit inside a cursor loop

From: Daniel Roy <danielroy10_at_hotmail.com>
Date: 30 May 2003 07:59:18 -0700
Message-ID: <1b061893.0305300659.5d90a9d1@posting.google.com>

Andy,
this is a very good procedure to avoid the risk of this ORA-1555. I once saw Tom Kyte give an example on how to do this, but I can't find it again. I think this note from metalink is very useful in understanding what really happens when we get an ORA-1555 error, and ways to avoid it:

Overview

This article will discuss the circumstances under which a query can return the
Oracle error ORA-01555 "snapshot too old (rollback segment too small)". The
article will then proceed to discuss actions that can be taken to avoid the
error and finally will provide some simple PL/SQL scripts that illustrate the
issues discussed.

Terminology

It is assumed that the reader is familiar with standard Oracle terminology such
as 'rollback segment' and 'SCN'. If not, the reader should first read the Oracle
Server Concepts manual and related Oracle documentation.

In addition to this, two key concepts are briefly covered below which help in
the understanding of ORA-01555:

READ CONSISTENCY:

This is documented in the Oracle Server Concepts manual and so will not be
discussed further. However, for the purposes of this article this should be read
and understood if not understood already.

Oracle Server has the ability to have multi-version read consistency which is
invaluable to you because it guarantees that you are seeing a consistent view of
the data (no 'dirty reads').

2. DELAYED BLOCK CLEANOUT:

This is best illustrated with an example: Consider a transaction that updates a
million row table. This obviously visits a large number of database blocks to
make the change to the data. When the user commits the transaction Oracle does
NOT go back and revisit these blocks to make the change permanent. It is left
for the next transaction that visits any block affected by the update to 'tidy
up' the block (hence the term 'delayed block cleanout').

Whenever Oracle changes a database block (index, table, cluster) it stores a
pointer in the header of the data block which identifies the rollback segment
used to hold the rollback information for the changes made by the transaction.
(This is required if the user later elects to not commit the changes and wishes
to 'undo' the changes made.)

Upon commit, the database simply marks the relevant rollback segment header
entry as committed. Now, when one of the changed blocks is revisited Oracle
examines the header of the data block which indicates that it has been changed
at some point. The database needs to confirm whether the change has been
committed or whether it is currently uncommitted. To do this, Oracle determines
the rollback segment used for the previous transaction (from the block's header)
and then determines whether the rollback header indicates whether it has been
committed or not.

If it is found that the block is committed then the header of the data block is
updated so that subsequent accesses to the block do not incur this processing.

This behaviour is illustrated in a very simplified way below. Here we walk
through the stages involved in updating a data block.

STAGE 1 - No changes made

Description: This is the starting point. At the top of the

              data block we have an area used to link active
              transactions to a rollback
              segment (the 'tx' part), and the rollback segment 
              header has a table that stores information upon
              all the latest transactions
              that have used that rollback segment. 

              In our example, we have two active transaction 
              slots (01 and 02)
              and the next free slot is slot 03. (Since we are 
              free to overwrite committed transactions.)

Data Block 500             Rollback Segment Header 5
+----+--------------+      +----------------------+---------+


| tx | None         |      | transaction entry 01 |ACTIVE   |

+----+--------------+      | transaction entry 02 |ACTIVE   |


| row 1             |      | transaction entry 03 |COMMITTED|

| row 2             |      | transaction entry 04 |COMMITTED|

| ... ..            |      |     ...     ...   .. |  ...    |

| row n             |      | transaction entry nn |COMMITTED|

+-------------------+       +--------------------------------+

STAGE 2 - Row 2 is updated

Description: We have now updated row 2 of block 500. Note that

              the data block header is updated to point to the
              rollback segment 5, transaction
              slot 3 (5.3) and that it is marked uncommitted 
             (Active). 

Data Block 500             Rollback Segment Header 5
+----+--------------+     +----------------------+---------+


| tx |5.3uncommitted|-+   | transaction entry 01 |ACTIVE   |

+----+--------------+ |   | transaction entry 02 |ACTIVE   |


| row 1             | +-->| transaction entry 03 |ACTIVE   |

| row 2 *changed*   |     | transaction entry 04 |COMMITTED|

| ... ..            |     |     ...     ...   .. |  ...    |

| row n             |     | transaction entry nn |COMMITTED|

+------------------+      +--------------------------------+

STAGE 3 - The user issues a commit

Description: Next the user hits commit. Note that all that

              this does is it
              updates the rollback segment header's 
              corresponding transaction
              slot as committed. It does *nothing* to the data 
              block.

Data Block 500                   Rollback Segment Header 5
+----+--------------+       +----------------------+---------+


| tx |5.3uncommitted|--+    | transaction entry 01 |ACTIVE   |

+----+--------------+  |    | transaction entry 02 |ACTIVE   |


| row 1             |  +--->| transaction entry 03 |COMMITTED|

| row 2 *changed*   |       | transaction entry 04 |COMMITTED|

| ... ..            |       |     ...     ...   .. |  ...    |

| row n             |       | transaction entry nn |COMMITTED|

+------------------+        +--------------------------------+

STAGE 4 - Another user selects data block 500

Description: Some time later another user (or the same user)

              revisits data block 500. We can see that there 
              is an uncommitted change in the
              data block according to the data block's header.

              Oracle then uses the data block header to look up
              the corresponding rollback segment transaction
              table slot, sees that it has been committed, and
              changes data block 500 to reflect the
              true state of the datablock. (i.e. it performs
              delayed cleanout).

Data Block 500                   Rollback Segment Header 5
+----+--------------+      +----------------------+---------+


| tx | None         |      | transaction entry 01 |ACTIVE   |

+----+--------------+      | transaction entry 02 |ACTIVE   |


| row 1             |      | transaction entry 03 |COMMITTED|

| row 2             |      | transaction entry 04 |COMMITTED|

| ... ..            |      |     ...     ...   .. |  ...   |

| row n             |      | transaction entry nn |COMMITTED|

+------------------+       +--------------------------------+

ORA-01555 Explanation

There are two fundamental causes of the error ORA-01555 that are a result of
Oracle trying to attain a 'read consistent' image. These are :

o The rollback information itself is overwritten so that Oracle is unable to
rollback the (committed) transaction entries to attain a sufficiently old enough
version of the block.

o The transaction slot in the rollback segment's transaction table (stored in
the rollback segment's header) is overwritten, and Oracle cannot rollback the
transaction header sufficiently to derive the original rollback segment
transaction slot.

Both of these situations are discussed below with the series of steps that cause
the ORA-01555. In the steps, reference is made to 'QENV'. 'QENV' is short for
'Query Environment', which can be thought of as the environment that existed
when a query is first started and to which Oracle is trying to attain a read
consistent image. Associated with this environment is the SCN (System Change Number) at that time and hence, QENV 50 is the query environment
with SCN 50.

CASE 1 - ROLLBACK OVERWRITTEN This breaks down into two cases: another session overwriting the rollback that
the current session requires or the case where the current session overwrites
the rollback information that it requires. The latter is discussed in this
article because this is usually the harder one to understand.

Steps:

Session 1 starts query at time T1 and QENV 50
Session 1 selects block B1 during this query
Session 1 updates the block at SCN 51
Session 1 does some other work that generates rollback information.
Session 1 commits the changes made in steps '3' and '4'. (Now other transactions are free to overwrite this rollback information)
Session 1 revisits the same block B1 (perhaps for a different row).

Now, Oracle can see from the block's header that it has been changed and
it is later than the required QENV (which was 50). Therefore we need to get an
image of the block as of this QENV.

If an old enough version of the block can be found in the buffer cache
then we will use this, otherwise we need to rollback the current block to
generate another version of the block as at the required QENV.

It is under this condition that Oracle may not be able to get the
required rollback information because Session 1's changes have generated
rollback information that has overwritten it and returns the ORA-1555 error.

CASE 2 - ROLLBACK TRANSACTION SLOT OVERWRITTEN

Session 1 starts query at time T1 and QENV 50
Session 1 selects block B1 during this query
Session 1 updates the block at SCN 51
Session 1 commits the changes (Now other transactions are free to overwrite this rollback information)
A session (Session 1, another session or a number of other sessions) then use the same rollback segment for a series of committed transactions.

These transactions each consume a slot in the rollback segment transaction table such that it eventually wraps around (the slots are written to
in a circular fashion) and overwrites all the slots. Note that Oracle is free to
reuse these slots since all transactions are committed.

6. Session 1's query then visits a block that has been changed since the
initial QENV was established. Oracle therefore needs to derive an image of the
block as at that point in time.

Next Oracle attempts to lookup the rollback segment header's transaction
slot pointed to by the top of the data block. It then realises that this has
been overwritten and attempts to rollback the changes made to the rollback
segment header to get the original transaction slot entry.

If it cannot rollback the rollback segment transaction table sufficiently
it will return ORA-1555 since Oracle can no longer derive the required version
of the data block.

It is also possible to encounter a variant of the transaction slot being
overwritten when using block cleanout. This is briefly described below :

Session 1 starts a query at QENV 50. After this another process updates the
blocks that Session 1 will require. When Session 1 encounters these blocks it
determines that the blocks have changed and have not yet been cleaned out (via
delayed block cleanout). Session 1 must determine whether the rows in the block
existed at QENV 50, were subsequently changed,

In order to do this, Oracle must look at the relevant rollback segment
transaction table slot to determine the committed SCN. If this SCN is after the
QENV then Oracle must try to construct an older version of the block and if it
is before then the block just needs clean out to be good enough for the QENV.

If the transaction slot has been overwritten and the transaction table cannot
be rolled back to a sufficiently old enough version then Oracle cannot derive
the block image and will return ORA-1555.

(Note: Normally Oracle can use an algorithm for determining a block's SCN
during block cleanout even when the rollback segment slot has been overwritten.
But in this case Oracle cannot guarantee that the version of the block has not
changed since the start of the query).

Solutions

This section lists some of the solutions that can be used to avoid the ORA-01555
problems discussed in this article. It addresses the cases where rollback
segment information is overwritten by the same session and when the rollback
segment transaction table entry is overwritten.

It is worth highlighting that if a single session experiences the ORA-01555 and
it is not one of the special cases listed at the end of this article, then the
session must be using an Oracle extension whereby fetches across commits are
tolerated. This does not follow the ANSI model and in the rare cases where
ORA-01555 is returned one of the solutions below must be used.

CASE 1 - ROLLBACK OVERWRITTEN

Increase size of rollback segment which will reduce the likelihood of overwriting rollback information that is needed.
Reduce the number of commits (same reason as 1).
Run the processing against a range of data rather than the whole table. (Same reason as 1).
Add additional rollback segments. This will allow the updates etc. to be spread across more rollback segments thereby reducing the chances of overwriting required rollback information.
If fetching across commits, the code can be changed so that this is not done.
Ensure that the outer select does not revisit the same block at different times during the processing. This can be achieved by :
- Using a full table scan rather than an index lookup
- Introducing a dummy sort so that we retrieve all the data, sort it and then sequentially visit these data blocks.

CASE 2 - ROLLBACK TRANSACTION SLOT OVERWRITTEN

Use any of the methods outlined above except for '6'. This will allow transactions to spread their work across multiple rollback segments therefore reducing the likelihood or rollback segment transaction table slots being consumed.
If it is suspected that the block cleanout variant is the cause, then force block cleanout to occur prior to the transaction that returns the ORA-1555. This can be achieved by issuing the following in SQL*Plus, SQL*DBA or Server Manager :

      alter session set optimizer_goal = rule;
      select count(*) from table_name;

     If indexes are being accessed then the problem may be an index

block and
clean out can be forced by ensuring that all the index is traversed. Eg, if the
index is on a numeric column with a minimum value of 25 then the following query
will force cleanout of the index :

select index_column from table_name where index_column > 24;

Examples

Listed below are some PL/SQL examples that can be used to illustrate the
ORA-1555 cases given above. Before these PL/SQL examples will return this error
the database must be configured as follows :

o Use a small buffer cache (db_block_buffers).

REASON: You do not want the session executing the script to be able to find
old versions of the block in the buffer cache which can be used to satisfy a
block visit without requiring the rollback information.

o Use one rollback segment other than SYSTEM.

REASON: You need to ensure that the work being done is generating rollback
information that will overwrite the rollback information required.

o Ensure that the rollback segment is small.

REASON: See the reason for using one rollback segment.

ROLLBACK OVERWRITTEN

rem * 1555_a.sql - 
rem * Example of getting ora-1555 "Snapshot too old" by
rem * session overwriting the rollback information required
rem * by the same session.

drop table bigemp;
create table bigemp (a number, b varchar2(30), done char(1));

drop table dummy1;
create table dummy1 (a varchar2(200));

  rem * Populate the example tables.
  begin
   for i in 1..4000 loop

     insert into bigemp values (mod(i,20), to_char(i), 'N');
     if mod(i,100) = 0 then
       insert into dummy1 values ('ssssssssssss');
       commit;
     end if;

   end loop;
   commit;
  end;
  /

rem * Ensure that table is 'cleaned out'. select count(*) from bigemp;

declare

Must use a predicate so that we revisit a changed block at a different
time.
If another tx is updating the table then we may not need the predicate cursor c1 is select rowid, bigemp.* from bigemp where a < 20;

begin
for c1rec in c1 loop

     update dummy1 set a = 'aaaaaaaa';
     update dummy1 set a = 'bbbbbbbb';
     update dummy1 set a = 'cccccccc';
     update bigemp set done='Y' where c1rec.rowid = rowid;
     commit;

   end loop;
  end;
  /

ROLLBACK TRANSACTION SLOT OVERWRITTEN

  rem * 1555_b.sql - Example of getting ora-1555 "Snapshot too old" by
  rem *              overwriting the transaction slot in the rollback
  rem *              segment header. This just uses one session.

drop table bigemp;
create table bigemp (a number, b varchar2(30), done char(1));

  rem * Populate demo table.
  begin
   for i in 1..200 loop

     insert into bigemp values (mod(i,20), to_char(i), 'N');
     if mod(i,100) = 0 then
       commit;
     end if;

   end loop;
   commit;
  end;
  /

  drop table mydual;
  create table mydual (a number);
  insert into mydual values (1);
  commit;

rem * Cleanout demo table.
select count(*) from bigemp;

declare

cursor c1 is select * from bigemp;

begin

The following update is required to illustrate the problem if block
cleanout has been done on 'bigemp'. If the cleanout (above) is commented
out then the update and commit statements can be commented and the
script will fail with ORA-1555 for the block cleanout variant. update bigemp set b = 'aaaaa'; commit;

for c1rec in c1 loop

     for i in 1..20 loop
       update mydual set a=a;
       commit;
     end loop;

   end loop;
  end;
  /

Special Cases

There are other special cases that may result in an ORA-01555. These are given
below but are rare and so not discussed in this article :

o Trusted Oracle can return this if configured in OS MAC mode. Decreasing
LOG_CHECKPOINT_INTERVAL on the secondary database may overcome the problem.

o If a query visits a data block that has been changed by using the Oracle
discrete transaction facility then it will return ORA-01555.

o It is feasible that a rollback segment created with the OPTIMAL clause
maycause a query to return ORA-01555 if it has shrunk during the life of the
query causing rollback segment information required to generate consistent read
versions of blocks to be lost.

Summary

This article has discussed the reasons behind the error ORA-01555 "Snapshot too
old", has provided a list of possible methods to avoid the error when it is
encountered, and has provided simple PL/SQL scripts that illustrate the cases
discussed.

Amen

Daniel Received on Fri May 30 2003 - 09:59:18 CDT