OC 2 Redo and Undo

Addenda and Errata for Oracle Core Chapter 2 Redo and Undo

Back to Index

Addenda

n/a	Here’s a link to an article I wrote some time ago about undo segments growing to extreme sizes. It’s a useful example demonstrating an important feature of how undo segments work.
p.9	Paragraph before heading Debriefing, I state: “The most significant detail, for our purposes, is the DBA: (data block address) in line 1, which identifies block 0x0080009a: if you know your Oracle block numbers in hex, you’ll recognize that this is block 154 of data file 2.”I should, perhaps, have written a note in the appendix about how to work this out and directed the reader to the appendix at this point. Here, however, is a method for working out the necessary detail: select dbms_utility.data_block_address_file(to_number('0080009a','XXXXXXXX')) file_no, dbms_utility.data_block_address_block(to_number('0080009a','XXXXXXXX')) block_no from dual ; Informally, however, and only for simple database with a small number (less than 64) of relatively small (less than 1M blocks) data files, you can often break a database block address into two parts based on the fact that the block component is the low 22 bits, with the file number mangled into the higher bits. If you know that your data files are less than 1M blocks, then the last 5 bytes of the data block address will be the block number within file. (0xFFFFF = 1,048,575). If you know the number of data files is less than 64 then the first three bytes divided 4 (i.e. right-shifted twice) is the file number. In this example we get: block = 0x9a = 154, and file = 0x008 / 4 = 2. It’s a quick and dirty algorithm for small test systems – but breaks when the file number goes above 63 because of the encoding that Oracle introduced as the maximum number of datafiles increased in the change from v6 to v7 (and again from v7 to v8), and gets messy when you have to allow for the top 2 bits when the number of blocks is so large that it runs into byte that is shared with the file number.
p.19	Last two lines: “The advantage of keeping undo information inside the database in “ordinary” data files is that the blocks are subject to …”. Removing a little ambiguity, this looks better as: “The advantage of keeping undo information inside the database in “ordinary” data files is that the undo blocks are subject to …”
p.21	Paragraph after Note, eighth line: “… the record pointed at by the irb: entry.” There is no previous explanation, or even mention, of the irb entry; this is an item that appears in the header section of an undo block (see page 32 for an extract from a block dump). The header section contains a count (cnt) of the number of records in the block, but also contains a pointer (irb – index to roll back ?) to which record should be used first in the event of a rollback call being made to terminate, or clean up, a transaction.There’s a brief mention of the irb (and Christian Antognini’s correction to my initial understanding of its meaning) at this URL.

Errata

p.13	In the Sidebar on Isolation levels lines 1 and 4 read “select from t1” – this wasn’t intended as a code example, but technically you would have to select something if you don’t want the SQL to fail. In a new release I might put: “select {list of columns} from t1”

Typos/Grammar/Style/Punctuation

p.6	End of first paragraph: “there are”, style change“the steps are:”
p.12	I didn’t give the table of ACID requirements a heading, so the editor has used the first requirement (atomicity) as if it were the heading.

Back to Index