Database design for audit logging [closed]_问答_开发者

Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.

We don’t allow questions seeking recommendations for books, tools, software libraries, and more. You can edit the question so it can be answered with facts and citations.

Closed last year.

The community reviewed whether to reopen this question last year and left it closed:

开发者_运维问答

Original close reason(s) were not resolved

Improve this question

Every time I need to design a new database I spend quite some time thinking on how I should set up the database schema to keep an audit log of the changes.

Some questions have already been asked here about this, but I don't agree that there is a single best approach for all scenarios:

Database Design For Revisions
Best design for a change log auditing database table
Ideas on database design for capturing audit trails

I have also stumbled upon this interesting article on Maintaining a Log of Database Changes that tries to list the pro and cons of each approach. It's very well written and has interesting information, but it has made my decisions even harder.

My question is: Is there a reference that I can use, maybe a book or something like a decision tree that I can refer to decide which way should I go based on some input variables, like:

The maturity of the database schema
How the logs will be queried
The probability that it will be need to recreate records
What's more important: write or read performance
Nature of the values that are being logged (string, numbers, blobs)
Storage space available

The approaches that I know are:

1. Add columns for created and modified date and user

Table example:

id
value_1
value_2
value_3
created_date
modified_date
created_by
modified_by

Major cons: We lose the history of the modifications. Can't rollback after commit.

2. Insert only tables

Table example:

id
value_1
value_2
value_3
from
to
deleted (Boolean)
user

Major cons: How to keep foreign keys up to date? Huge space needed

3. Create a Separate history table for each table

History table example:

id
value_1
value_2
value_3
value_4
user
deleted (Boolean)
timestamp

Major cons: Needs to duplicate all audited tables. If the schema changes it will be needed to the migrate all the logs too.

4. Create a Consolidated history Table for All Tables

History table example:

table_name
field
user
new_value
deleted (Boolean)
timestamp

Major cons: Will I be able to recreate the records (rollback) if needed easily? The new_value column needs to be a huge string so it can support all different column types.

One method that is used by a few wiki platforms is to separate the identifying data and the content you're auditing. It adds complexity, but you end up with an audit trail of complete records, not just listings of fields that were edited that you then have to mash up to give the user an idea of what the old record looked like.

So for example, if you had a table called Opportunities to track sales deals, you would actually create two separate tables:

Opportunities
Opportunities_Content (or something like that)

The Opportunities table would have information you'd use to uniquely identify the record and would house the primary key you'd reference for your foreign key relationships. The Opportunities_Content table would hold all the fields your users can change and for which you'd like to keep an audit trail. Each record in the Content table would include its own PK and the modified-by and modified-date data. The Opportunities table would include a reference to the current version as well as information on when the main record was originally created and by whom.

Here's a simple example:

CREATE TABLE dbo.Page(  
    ID int PRIMARY KEY,  
    Name nvarchar(200) NOT NULL,  
    CreatedByName nvarchar(100) NOT NULL, 
    CurrentRevision int NOT NULL, 
    CreatedDateTime datetime NOT NULL

And the contents:

CREATE TABLE dbo.PageContent(
    PageID int NOT NULL,
    Revision int NOT NULL,
    Title nvarchar(200) NOT NULL,
    User nvarchar(100) NOT NULL,
    LastModified datetime NOT NULL,
    Comment nvarchar(300) NULL,
    Content nvarchar(max) NOT NULL,
    Description nvarchar(200) NULL

I would probably make the PK of the contents table a multi-column key from PageID and Revision provided Revision was an identity type. You would use the Revision column as the FK. You then pull the consolidated record by JOINing like this:

SELECT * FROM Page
JOIN PageContent ON CurrentRevision = Revision AND ID = PageID

There might be some errors up there...this is off the top of my head. It should give you an idea of an alternative pattern, though.

If you are using SQL Server 2008, you probably should consider Change Data Capture. This is new for 2008 and could save you a considerable amount of work.

I don't know of any reference, but I'm sure someone has written something.

However, if the purpose is simply to have a record of what happened—the most typical use of an audit log—then why not simply keep everything:

timestamp
username
ip_address
procedureName (if called from a stored procedure)
database
table
field
accesstype (insert, delete, modify)
oldvalue
newvalue

Presumably this is maintained by a trigger.

We’ll create a small example database for a blogging application. Two tables are required:

blog: stores a unique post ID, the title, content, and a deleted flag. audit: stores a basic set of historical changes with a record ID, the blog post ID, the change type (NEW, EDIT or DELETE) and the date/time of that change. The following SQL creates the blog and indexes the deleted column:

CREATE TABLE `blog` (
    `id` mediumint(8) unsigned NOT NULL AUTO_INCREMENT,
    `title` text,
    `content` text,
    `deleted` tinyint(1) unsigned NOT NULL DEFAULT '0',
    PRIMARY KEY (`id`),
    KEY `ix_deleted` (`deleted`)
) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8 COMMENT='Blog posts';

The following SQL creates the audit table. All columns are indexed and a foreign key is defined for audit.blog_id which references blog.id. Therefore, when we physically DELETE a blog entry, it’s full audit history is also removed.

CREATE TABLE `audit` (
    `id` mediumint(8) unsigned NOT NULL AUTO_INCREMENT,
    `blog_id` mediumint(8) unsigned NOT NULL,
    `changetype` enum('NEW','EDIT','DELETE') NOT NULL,
    `changetime` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
    PRIMARY KEY (`id`),
    KEY `ix_blog_id` (`blog_id`),
    KEY `ix_changetype` (`changetype`),
    KEY `ix_changetime` (`changetime`),
    CONSTRAINT `FK_audit_blog_id` FOREIGN KEY (`blog_id`) REFERENCES `blog` (`id`) ON DELETE CASCADE ON UPDATE CASCADE
) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8;

I think there is nothing like a decision tree. Since some of the pros and cons (or the requirements) are not really countable. How do you measure Maturity for instance?

So just line up your business requirements for your audit logging. Try to predict how these requirements might change in the future and generate your technical requirements. Now you can compare it to the pros and cons and choose the right/best option.

And be assured, it doesn't matter how you decide, there will always someone who think you made the wrong decision. However, you did your homework and you justify your decision.

I'm using the following structure:

id  int
user_id int
system_user_id  int
tenant_id   int
db_name varchar
model_name  varchar
model_primary_key   int
model_attributes    text
created_at  timestamp
ip  varchar
session_id  varchar
request_id  varchar
comments    text

Works well so far with ~362 million records, multi-tenant, multi-database.

model_attributes is the most important, what changed, as json string in key value format.