Mozilla/mozilla/webtools/bugzilla/docs/xml/database.xml

<!-- <!DOCTYPE appendix PUBLIC "-//OASIS//DTD DocBook V4.1//EN"> -->

<appendix id="database">

<title>The Bugzilla Database</title>
<note>
<para>
      This document really needs to be updated with more fleshed out information about primary keys, interrelationships, and maybe some nifty tables to document dependencies.  Any takers?
    </para>
  </note>
  <section id="dbschema">
    <title>Database Schema Chart</title>
    <para>
      <mediaobject>
	<imageobject>
	  <imagedata fileref="../images/dbschema.jpg" format="JPG" />
	</imageobject>
	
	<textobject>
	  <phrase>Database Relationships</phrase>
	</textobject>

	<caption>
	  <para>Bugzilla database relationships chart</para>
	</caption>
      </mediaobject>
    </para>
  </section>

  <section id="dbdoc">
<title>MySQL Bugzilla Database Introduction</title>
    <para>
      This information comes straight from my life.  I was forced to learn how
      Bugzilla organizes database because of nitpicky requests from users for tiny
      changes in wording, rather than having people re-educate themselves or
      figure out how to work our procedures around the tool.  It sucks, but it can
      and will happen to you, so learn how the schema works and deal with it when it
      comes.
    </para>

  <para>
      So, here you are with your brand-new installation of Bugzilla.  You've got
      MySQL set up, Apache working right, Perl DBI and DBD talking to the database
      flawlessly.  Maybe you've even entered a few test bugs to make sure email's
      working; people seem to be notified of new bugs and changes, and you can
      enter and edit bugs to your heart's content.  Perhaps you've gone through the
      trouble of setting up a gateway for people to submit bugs to your database via
      email, have had a few people test it, and received rave reviews from your beta
      testers.
    </para>
    <para>
      What's the next thing you do?  Outline a training strategy for your
      development team, of course, and bring them up to speed on the new tool you've
      labored over for hours.
    </para>
    <para>
      Your first training session starts off very well!  You have a captive
      audience which seems enraptured by the efficiency embodied in this thing called
      "Bugzilla".  You are caught up describing the nifty features, how people can
      save favorite queries in the database, set them up as headers and footers on
      their pages, customize their layouts, generate reports, track status with
      greater efficiency than ever before, leap tall buildings with a single bound
      and rescue Jane from the clutches of Certain Death!
    </para>
    <para>
      But Certain Death speaks up -- a tiny voice, from the dark corners of the
      conference room.  "I have a concern," the voice hisses from the darkness,
      "about the use of the word 'verified'.
    </para>
    <para>
      The room, previously filled with happy chatter, lapses into reverential
      silence as Certain Death (better known as the Vice President of Software
      Engineering) continues.  "You see, for two years we've used the word 'verified'
      to indicate that a developer or quality assurance engineer has confirmed that,
      in fact, a bug is valid. I don't want to lose two years of training to a
      new software product.  You need to change the bug status of 'verified' to
      'approved' as soon as possible. To avoid confusion, of course."
    </para>
    <para>
      Oh no!  Terror strikes your heart, as you find yourself mumbling "yes, yes, I
      don't think that would be a problem," You review the changes with Certain
      Death, and continue to jabber on, "no, it's not too big a change. I mean, we
      have the source code, right? You know, 'Use the Source, Luke' and all that...
      no problem," All the while you quiver inside like a beached jellyfish bubbling,
      burbling, and boiling on a hot Jamaican sand dune...
    </para>
    <para>
      Thus begins your adventure into the heart of Bugzilla.  You've been forced
      to learn about non-portable enum() fields, varchar columns, and tinyint
      definitions. The Adventure Awaits You!
    </para>

    <section>
      <title>Bugzilla Database Basics</title>
      <para>
	If you were like me, at this point you're totally clueless
	about the internals of MySQL, and if it weren't for this
	executive order from the Vice President you couldn't care less
	about the difference between a <quote>bigint</quote> and a
	<quote>tinyint</quote> entry in MySQL.   I recommend you refer
	to the MySQL documentation, available at <ulink url="http://www.mysql.com/doc.html">MySQL.com</ulink>.  Below are the basics you need to know about the Bugzilla database.  Check the chart above for more details.
      </para>
      <para><orderedlist>
	  <listitem>
	    <para>
	      To connect to your database:
	    </para>
	    <para>
	      <prompt>bash#</prompt><command>mysql</command><parameter>-u root</parameter>
	    </para>
	    <para>
	      If this works without asking you for a password,
	      <emphasis>shame on you</emphasis>!  You should have
	      locked your security down like the installation
	      instructions told you to.  You can find details on
	      locking down your database in the Bugzilla FAQ in this
	      directory (under "Security"), or more robust security
	      generalities in the MySQL searchable documentation at
	      http://www.mysql.com/php/manual.php3?section=Privilege_system .
	    </para>
	  </listitem>

	  <listitem>
	    <para>You should now be at a prompt that looks like
	      this:</para>
	    <para><prompt>mysql></prompt></para>
	    <para>At the prompt, if <quote>bugs</quote> is the name
	      you chose in the<filename>localconfig</filename> file
	      for your Bugzilla database, type:</para>
	    <para><prompt>mysql</prompt><command>use bugs;</command></para>
	    <note>
	      <para>Don't forget the <quote>;</quote> at the end of
		each line, or you'll be kicking yourself later.</para>
	    </note>
	  </listitem>
	</orderedlist>
      </para>
      <section>
	<title>Bugzilla Database Tables</title>
	<para>  Imagine your MySQL database as a series of
	  spreadsheets, and you won't be too far off.  If you use this
	  command:</para>
	<para><prompt>mysql></prompt><command>show tables from bugs;</command></para>
	<para>you'll be able to see all the
	  <quote>spreadsheets</quote> (tables) in your database.  It
	  is similar to a file system, only faster and more robust for
	  certain types of operations.</para>
	<para>From the command issued above, ou should have some
	  output that looks like this:
	  <programlisting>
+-------------------+
| Tables in bugs    |
+-------------------+
| attachments       |
| bugs              |
| bugs_activity     |
| cc                |
| components        |
| dependencies      |
| fielddefs         |
| groups            |
| keyworddefs       |
| keywords          |
| logincookies      |
| longdescs         |
| milestones        |
| namedqueries      |
| products          |
| profiles          |
| profiles_activity |
| shadowlog         |
| tokens            |
| versions          |
| votes             |
| watch             |
+-------------------+
	  </programlisting></para>
<literallayout>

  Here's an overview of what each table does.  Most columns in each table have
descriptive names that make it fairly trivial to figure out their jobs.

attachments: This table stores all attachments to bugs.  It tends to be your
largest table, yet also generally has the fewest entries because file
attachments are so (relatively) large.

bugs:  This is the core of your system.  The bugs table stores most of the
current information about a bug, with the exception of the info stored in the
other tables.

bugs_activity:  This stores information regarding what changes are made to bugs
when -- a history file.

cc:  This tiny table simply stores all the CC information for any bug which has
any entries in the CC field of the bug.  Note that, like most other tables in
Bugzilla, it does not refer to users by their user names, but by their unique
userid, stored as a primary key in the profiles table.

components: This stores the programs and components (or products and
components, in newer Bugzilla parlance) for Bugzilla.  Curiously, the "program"
(product) field is the full name of the product, rather than some other unique
identifier, like bug_id and user_id are elsewhere in the database.

dependencies: Stores data about those cool dependency trees.

fielddefs:  A nifty table that defines other tables.  For instance, when you
submit a form that changes the value of "AssignedTo" this table allows
translation to the actual field name "assigned_to" for entry into MySQL.

groups:  defines bitmasks for groups.  A bitmask is a number that can uniquely
identify group memberships.  For instance, say the group that is allowed to
tweak parameters is assigned a value of "1", the group that is allowed to edit
users is assigned a "2", and the group that is allowed to create new groups is
assigned the bitmask of "4".  By uniquely combining the group bitmasks (much
like the chmod command in UNIX,) you can identify a user is allowed to tweak
parameters and create groups, but not edit users, by giving him a bitmask of
"5", or a user allowed to edit users and create groups, but not tweak
parameters, by giving him a bitmask of "6" Simple, huh?
  If this makes no sense to you, try this at the mysql prompt:
mysql> select * from groups;
  You'll see the list, it makes much more sense that way.

keyworddefs:  Definitions of keywords to be used

keywords: Unlike what you'd think, this table holds which keywords are
associated with which bug id's.

logincookies: This stores every login cookie ever assigned to you for every
machine you've ever logged into Bugzilla from.  Curiously, it never does any
housecleaning -- I see cookies in this file I've not used for months.  However,
since Bugzilla never expires your cookie (for convenience' sake), it makes
sense.

longdescs:  The meat of bugzilla -- here is where all user comments are stored!
You've only got 2^24 bytes per comment (it's a mediumtext field), so speak
sparingly -- that's only the amount of space the Old Testament from the Bible
would take (uncompressed, 16 megabytes).  Each comment is keyed to the
bug_id to which it's attached, so the order is necessarily chronological, for
comments are played back in the order in which they are received.

milestones:  Interesting that milestones are associated with a specific product
in this table, but Bugzilla does not yet support differing milestones by
product through the standard configuration interfaces.

namedqueries:  This is where everybody stores their "custom queries".  Very
cool feature; it beats the tar out of having to bookmark each cool query you
construct.

products:  What products you have, whether new bug entries are allowed for the
product, what milestone you're working toward on that product, votes, etc.  It
will be nice when the components table supports these same features, so you
could close a particular component for bug entry without having to close an
entire product...

profiles:  Ahh, so you were wondering where your precious user information was
stored?  Here it is!  With the passwords in plain text for all to see! (but
sshh... don't tell your users!)

profiles_activity:  Need to know who did what when to who's profile?  This'll
tell you, it's a pretty complete history.

shadowlog:  I could be mistaken here, but I believe this table tells you when
your shadow database is updated and what commands were used to update it.  We
don't use a shadow database at our site yet, so it's pretty empty for us.

versions:  Version information for every product

votes:  Who voted for what when

watch:  Who (according to userid) is watching who's bugs (according to their
userid).


===
THE DETAILS
===

  Ahh, so you're wondering just what to do with the information above?  At the
mysql prompt, you can view any information about the columns in a table with
this command (where "table" is the name of the table you wish to view):

mysql> show columns from table;

  You can also view all the data in a table with this command:

mysql> select * from table;

  -- note: this is a very bad idea to do on, for instance, the "bugs" table if
you have 50,000 bugs.  You'll be sitting there a while until you ctrl-c or
50,000 bugs play across your screen.

  You can limit the display from above a little with the command, where
"column" is the name of the column for which you wish to restrict information:

mysql> select * from table where (column = "some info");

  -- or the reverse of this

mysql> select * from table where (column != "some info");

  Let's take our example from the introduction, and assume you need to change
the word "verified" to "approved" in the resolution field.  We know from the
above information that the resolution is likely to be stored in the "bugs"
table. Note we'll need to change a little perl code as well as this database
change, but I won't plunge into that in this document. Let's verify the
information is stored in the "bugs" table:

mysql> show columns from bugs

  (exceedingly long output truncated here)
| bug_status| enum('UNCONFIRMED','NEW','ASSIGNED','REOPENED','RESOLVED','VERIFIED','CLOSED')||MUL | UNCONFIRMED||

  Sorry about that long line.  We see from this that the "bug status" column is
an "enum field", which is a MySQL peculiarity where a string type field can
only have certain types of entries.  While I think this is very cool, it's not
standard SQL.  Anyway, we need to add the possible enum field entry
'APPROVED' by altering the "bugs" table.

mysql> ALTER table bugs CHANGE bug_status bug_status
    -> enum("UNCONFIRMED", "NEW", "ASSIGNED", "REOPENED", "RESOLVED",
    -> "VERIFIED", "APPROVED", "CLOSED") not null;

    (note we can take three lines or more -- whatever you put in before the
semicolon is evaluated as a single expression)

Now if you do this:

mysql> show columns from bugs;

  you'll see that the bug_status field has an extra "APPROVED" enum that's
available!  Cool thing, too, is that this is reflected on your query page as
well -- you can query by the new status.  But how's it fit into the existing
scheme of things?
  Looks like you need to go back and look for instances of the word "verified"
in the perl code for Bugzilla -- wherever you find "verified", change it to
"approved" and you're in business (make sure that's a case-insensitive search).
Although you can query by the enum field, you can't give something a status
of "APPROVED" until you make the perl changes.   Note that this change I
mentioned can also be done by editing checksetup.pl, which automates a lot of
this.  But you need to know this stuff anyway, right?

  I hope this database tutorial has been useful for you.  If you have comments
to add, questions, concerns, etc. please direct them to
mbarnson@excitehome.net.  Please direct flames to /dev/null :)  Have a nice
day!



===
LINKS
===

Great MySQL tutorial site:
http://www.devshed.com/Server_Side/MySQL/

	</literallayout>
      </section>
    </section>
  </section>

  <section id="granttables">
    <title>MySQL Permissions &amp; Grant Tables</title>
    
    <note>
      <para>The following portion of documentation comes from my
	answer to an old discussion of Keystone, a cool product that
	does trouble-ticket tracking for IT departments.  I wrote this
	post to the  Keystone support group regarding MySQL grant
	table permissions, and how to use them effectively. It is
	badly in need of updating, as I believe MySQL has added a
	field or two to the grant tables  since this time, but it
	serves as a decent introduction and troubleshooting document
	for grant table issues.  I used Keynote to track my troubles
	until I discovered Bugzilla, which gave me a whole new set of
	troubles to work on : )  Although it is of limited use, it
	still has SOME use, thus it's still included.</para>
      <para>
	Please note, however, that I was a relatively new user to
	MySQL at the time.  Some of my suggestions, particularly in
	how to set up security, showed a terrible lack of
	security-related database experience.
      </para>
    </note>
    
    <literallayout>
From matt_barnson@singletrac.com Wed Jul  7 09:00:07 1999
Date: Mon, 1 Mar 1999 21:37:04 -0700 
From: Matthew Barnson matt_barnson@singletrac.com
To: keystone-users@homeport.org
Subject: [keystone-users] Grant Tables FAQ

    [The following text is in the "iso-8859-1" character set]
    [Your display is set for the "US-ASCII" character set]
    [Some characters may be displayed incorrectly]

Maybe we can include this rambling message in the Keystone FAQ?  It gets
asked a lot, and the only option current listed in the FAQ is
"--skip-grant-tables".

Really, you can't go wrong by reading section 6 of the MySQL manual, at
http://www.mysql.com/Manual/manual.html.  I am sure their description is
better than mine.

MySQL runs fine without permissions set up correctly if you run the mysql
daemon with the "--skip-grant-tables" option.  Running this way denies
access to nobody.  Unfortunately, unless you've got yourself firewalled it
also opens the potential for abuse if someone knows you're running it.

Additionally, the default permissions for MySQL allow anyone at localhost
access to the database if the database name begins with "test_" or is named
"test" (i.e. "test_keystone").  You can change the name of your database in
the keystone.conf file ($sys_dbname).  This is the way I am doing it for
some of my databases, and it works fine.

The methods described below assume you're running MySQL on the same box as
your webserver, and that you don't mind if your $sys_dbuser for Keystone has
superuser access.  See near the bottom of this message for a description of
what each field does.

Method #1:

1.  cd /var/lib
	#location where you'll want to run /usr/bin/mysql_install_db shell
script from to get it to work.

2.  ln -s mysql data  
	# soft links the "mysql" directory to "data", which is what
mysql_install_db expects.  Alternately, you can edit mysql_install_db and
change all the "./data" references to "./mysql".

3.  Edit /usr/bin/mysql_install_db with your favorite text editor (vi,
emacs, jot, pico, etc.)
A)  Copy the "INSERT INTO db VALUES
('%','test\_%','','Y','Y','Y','Y','Y','Y');" and paste it immediately after
itself.  Chage the 'test\_%' value to 'keystone', or the value of
$sys_dbname in keystone.conf.
B)  If you are running your keystone database with any user, you'll need to
copy the "INSERT INTO user VALUES
('localhost','root','','Y','Y','Y','Y','Y','Y','Y','Y','Y','Y');" line after
itself and change 'root' to the name of the keystone database user
($sys_dbuser) in keystone.conf.

	# adds entries to the script to create grant tables for specific
hosts and users.  The user you set up has super-user access ($sys_dbuser) --
you may or may not want this.  The layout of mysql_install_db is really very
uncomplicated.

4.  /usr/bin/mysqladmin shutdown
	# ya gotta shut it down before you can reinstall the grant tables!

5.  rm -i /var/lib/mysql/mysql/*.IS?' and answer 'Y' to the deletion
questions.
	# nuke your current grant tables.  This WILL NOT delete any other
databases than your grant tables.

6.  /usr/bin/mysql_install_db
	# run the script you just edited to install your new grant tables.

7.  mysqladmin -u root password (new_password)  
	# change the root MySQL password, or else anyone on localhost can
login to MySQL as root and make changes.  You can skip this step if you want
keystone to connect as root with no password.

8.  mysqladmin -u (webserver_user_name) password (new_password)  
	# change the password of the $sys_dbuser.  Note that you will need
to change the password in the keystone.conf file as well in $sys_dbpasswd,
and if your permissions are set up incorrectly anybody can type the URL to
your keystone.conf file and get the password.  Not that this will help them
much if your permissions are set to @localhost.



Method #2:  easier, but a pain reproducing if you have to delete your grant
tables.  This is the "recommended" method for altering grant tables in
MySQL.  I don't use it because I like the other way :)

shell> mysql --user=root keystone

mysql> GRANT
SELECT,INSERT,UPDATE,DELETE,INDEX,ALTER,CREATE,DROP,RELOAD,SHUTDOWN,PROCESS,
FILE,
           ON keystone.*
           TO &lt;$sys_dbuser name>@localhost
           IDENTIFIED BY '(password)'
	     WITH GRANT OPTION;

OR

mysql> GRANT ALL PRIVELEGES 
		ON keystone.*
		TO &lt;$sys_dbuser name>@localhost
		IDENTIFIED BY '(password)'
		WITH GRANT OPTION;

	# this grants the required permissions to the keystone ($sys_dbuser)
account defined in keystone.conf.  However, if you are runnning many
different MySQL-based apps, as we are, it's generally better to edit the
mysql_install_db script to be able to quickly reproduce your permissions
structure again.  Note that the FILE privelege and WITH GRANT OPTION may not
be in your best interest to include.


GRANT TABLE FIELDS EXPLANATION:
Quick syntax summary:  "%" in MySQL is a wildcard.  I.E., if you are
defining your DB table and in the 'host' field and enter '%', that means
that any host can access that database.  Of course, that host must also have
a valid db user in order to do anything useful.  'db'=name of database.  In
our case, it should be "keystone".  "user" should be your "$sys_dbuser"
defined in keystone.conf.  Note that you CANNOT add or change a password by
using the "INSERT INTO db (X)" command -- you must change it with the mysql
-u command as defined above.  Passwords are stored encrypted in the MySQL
database, and if you try to enter it directly into the table they will not
match.

TABLE:  USER.  Everything after "password" is a privelege granted (Y/N).
This table controls individual user global access rights.

'host','user','password','select','insert','update','delete','index','alter'
,'create','drop','grant','reload','shutdown','process','file'

TABLE:  DB.  This controls access of USERS to databases.

'host','db','user','select','insert','update','delete','index','alter','crea
te','drop','grant'

TABLE:  HOST.  This controls which HOSTS are allowed what global access
rights.  Note that the HOST table, USER table, and DB table are very closely
connected -- if an authorized USER attempts an SQL request from an
unauthorized HOST, she's denied.  If a request from an authorized HOST is
not an authorized USER, it is denied.  If a globally authorized USER does
not have rights to a certain DB, she's denied.  Get the picture?

'host','db','select','insert','update','delete','index','alter','create','dr
op','grant'


You should now have a working knowledge of MySQL grant tables.  If there is
anything I've left out of this answer that you feel is pertinent, or if my
instructions don't work for you, please let me know and I'll re-post this
letter again, corrected.  I threw it together one night out of exasperation
for all the newbies who don't know squat about MySQL yet, so it is almost
guaranteed to have errors.

Once again, you can't go wrong by reading section 6 of the MySQL manual.  It
is more detailed than I!
http://www.mysql.com/Manual/manual.html.

    </literallayout>
  </section>
  
</appendix>

<!-- Keep this comment at the end of the file
Local variables:
mode: sgml
sgml-always-quote-attributes:t
sgml-auto-insert-required-elements:t
sgml-balanced-tag-edit:t
sgml-exposed-tags:nil
sgml-general-insert-case:lower
sgml-indent-data:t
sgml-indent-step:2
sgml-local-catalogs:nil
sgml-local-ecat-files:nil
sgml-minimize-attributes:nil
sgml-namecase-general:t
sgml-omittag:t
sgml-parent-document:("Bugzilla-Guide.sgml" "book" "chapter")
sgml-shorttag:t
sgml-tag-region-if-active:t
End:
-->