Infrastructure

Jonathan Lewis's picture

Describe Upgrade

Here’s an odd little change between Oracle versions that could have a stunning impact on the application performance if the thing that generates your client code happens to use an unlucky selection of constructs.  It’s possible to demonstrate the effect remarkably easily – you just have to describe a table, doing it lots of times to make it easy to see what’s happening.

Jonathan Lewis's picture

sys_op_lbid

I’ve made use of the function a few times in the past, for example in this posting on the dangers of using reverse key indexes, but every time I’ve mentioned it I’ve only been interested in the “leaf blocks per key” option. There are actually four different variations of the function, relevant to different types of index and controlled by setting a flag parameter to one of 4 different values.

The call to sys_op_lbid() take 3 parameters: index (or index [sub]partition object id, a flag vlaue, and a table “rowid”, where the flag value can be one of L, R, O, or G. The variations of the call are as follows:

Jonathan Lewis's picture

Append hint

One of the questions that came up on the CBO Panel Session at the UKOUG Tech2018 conference was about the /*+ append */ hint – specifically how to make sure it was ignored when it came from a 3rd party tool that was used to load data into the database. The presence of the hint resulted in increasing amounts of space in the table being “lost” as older data was deleted by the application which then didn’t reuse the space the inserts always went above the table’s highwater mark; and it wasn’t possible to change the application code.

The first suggestion aired was to create an SQL Patch to associate the hint /*+ ignore_optim_embedded_hints */ with the SQL in the hope that this would make Oracle ignore the append hint. This won’t work, of course, because the append hint is not an optimizer hint, it’s a “behaviour” hint.

Jonathan Lewis's picture

DML Tablescans

This note is a follow-up to a recent comment a blog note about Row Migration:

So I wonder what is the difference between the two, parallel dml and serial dml with parallel scan, which makes them behave differently while working with migrated rows. Why might the strategy of serial dml with parallel scan case not work in parallel dml case? I am going to make a service request to get some clarifications but maybe I miss something obvious?

The comment also referenced a couple of MoS notes:

Jonathan Lewis's picture

Row Migration

There’s a little detail of row migration that’s been bugging me for a long time – and I’ve finally found a comment on MoS explaining why it happens. Before saying anything, though, else I’m going to give you a little script (that I’ve run on 12.2.0.1 with an 8KB block size in a tablespace using [corrected ASSM]  manual (freelist) space management and system allocated extents) to demonstrate the anomaly.

Jonathan Lewis's picture

Counting Rows

Here’s another little utility I use from time to time (usually for small tables) to check how many rows there are in each block of the table, and which blocks are used. It doesn’t do anything clever, just call routines in the dbms_rowid package for each rowid in the table:

Jonathan Lewis's picture

Dump logfile

Here’s a little procedure I’ve been using since Oracle 8i to dump the contents of the current log file – I’ve mentioned it several times in the past but never published it, so I’ll be checking for references to it and linking to it.

The code hasn’t changed in a long time, although I did add a query to get the full tracefile name from v$process when that became available. There’s also an (optional) called to dbms_support.my_sid to pick up the SID of the current session that slid into the code when that package became available.

Jonathan Lewis's picture

Index Splits – 3

This is stored only for reference, and in case anyone wants to wade through the details. It’s the redo log dump from the 90/10 index leaf block split test from the previous blog posts running on 11.2.0.4 on Linux. The first part is the full block dump, the second part is an extract of the Record and Change vector headings with the embedded opcode (opc:) for the undo records in the redo vectors, and a tiny note of what each change vector is doing.

Jonathan Lewis's picture

Index Splits – 2

In yesterday’s article I described the mechanism that Oracle for an index leaf block split when you try to insert a new entry into a leaf block that is already full, and I demonstrated that the “50-50” split and the “90-10” split work in the same way, namely:

Jonathan Lewis's picture

Index splits

After writing this note I came to the conclusion that it will be of no practical benefit to anyone …  but I’m publishing it anyway because it’s just an interesting little observation about the thought processes of some Oracle designer/developer. (Or maybe it’s an indication of how it’s sensible to re-use existing code rather than coding for a particular boundary case, or maybe it’s an example of how to take advantage of “dead time” to add a little icing to the cake when the extra time required might not get noticed). Anyway, the topic came up in a recent thread on the OTN/ODC database forum and since the description given there wasn’t quite right I thought I’d write up a correction and a few extra notes.

To prevent automated spam submissions leave this field empty.
Syndicate content