1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
|
Realm configuration decisions
=============================
Before installing Kerberos V5, it is necessary to consider the
following issues:
* The name of your Kerberos realm (or the name of each realm, if you
need more than one).
* How you will assign your hostnames to Kerberos realms.
* Which ports your KDC and and kadmind services will use, if they will
not be using the default ports.
* How many slave KDCs you need and where they should be located.
* The hostnames of your master and slave KDCs.
* How frequently you will propagate the database from the master KDC
to the slave KDCs.
Realm name
----------
Although your Kerberos realm can be any ASCII string, convention is to
make it the same as your domain name, in upper-case letters.
For example, hosts in the domain ``example.com`` would be in the
Kerberos realm::
EXAMPLE.COM
If you need multiple Kerberos realms, MIT recommends that you use
descriptive names which end with your domain name, such as::
BOSTON.EXAMPLE.COM
HOUSTON.EXAMPLE.COM
.. _mapping_hostnames:
Mapping hostnames onto Kerberos realms
--------------------------------------
Mapping hostnames onto Kerberos realms is done in one of three ways.
The first mechanism works through a set of rules in the
:ref:`domain_realm` section of :ref:`krb5.conf(5)`. You can specify
mappings for an entire domain or on a per-hostname basis. Typically
you would do this by specifying the mappings for a given domain or
subdomain and listing the exceptions.
The second mechanism is to use KDC host-based service referrals. With
this method, the KDC's krb5.conf has a full [domain_realm] mapping for
hosts, but the clients do not, or have mappings for only a subset of
the hosts they might contact. When a client needs to contact a server
host for which it has no mapping, it will ask the client realm's KDC
for the service ticket, and will receive a referral to the appropriate
service realm.
To use referrals, clients must be running MIT krb5 1.6 or later, and
the KDC must be running MIT krb5 1.7 or later. The
**host_based_services** and **no_host_referral** variables in the
:ref:`kdc_realms` section of :ref:`kdc.conf(5)` can be used to
fine-tune referral behavior on the KDC.
It is also possible for clients to use DNS TXT records, if
**dns_lookup_realm** is enabled in :ref:`krb5.conf(5)`. Such lookups
are disabled by default because DNS is an insecure protocol and security
holes could result if DNS records are spoofed. If enabled, the client
will try to look up a TXT record formed by prepending the prefix
``_kerberos`` to the hostname in question. If that record is not
found, the client will attempt a lookup by prepending ``_kerberos`` to the
host's domain name, then its parent domain, up to the top-level domain.
For the hostname ``boston.engineering.example.com``, the names looked up
would be::
_kerberos.boston.engineering.example.com
_kerberos.engineering.example.com
_kerberos.example.com
_kerberos.com
The value of the first TXT record found is taken as the realm name.
Even if you do not choose to use this mechanism within your site,
you may wish to set it up anyway, for use when interacting with other sites.
Ports for the KDC and admin services
------------------------------------
The default ports used by Kerberos are port 88 for the KDC and port
749 for the admin server. You can, however, choose to run on other
ports, as long as they are specified in each host's
:ref:`krb5.conf(5)` files or in DNS SRV records, and the
:ref:`kdc.conf(5)` file on each KDC. For a more thorough treatment of
port numbers used by the Kerberos V5 programs, refer to the
:ref:`conf_firewall`.
Slave KDCs
----------
Slave KDCs provide an additional source of Kerberos ticket-granting
services in the event of inaccessibility of the master KDC. The
number of slave KDCs you need and the decision of where to place them,
both physically and logically, depends on the specifics of your
network.
Kerberos authentication requires that each client be able to contact a
KDC. Therefore, you need to anticipate any likely reason a KDC might
be unavailable and have a slave KDC to take up the slack.
Some considerations include:
* Have at least one slave KDC as a backup, for when the master KDC is
down, is being upgraded, or is otherwise unavailable.
* If your network is split such that a network outage is likely to
cause a network partition (some segment or segments of the network
to become cut off or isolated from other segments), have a slave KDC
accessible to each segment.
* If possible, have at least one slave KDC in a different building
from the master, in case of power outages, fires, or other localized
disasters.
.. _kdc_hostnames:
Hostnames for KDCs
------------------
MIT recommends that your KDCs have a predefined set of CNAME records
(DNS hostname aliases), such as ``kerberos`` for the master KDC and
``kerberos-1``, ``kerberos-2``, ... for the slave KDCs. This way, if
you need to swap a machine, you only need to change a DNS entry,
rather than having to change hostnames.
As of MIT krb5 1.4, clients can locate a realm's KDCs through DNS
using SRV records (:rfc:`2782`), assuming the Kerberos realm name is
also a DNS domain name. These records indicate the hostname and port
number to contact for that service, optionally with weighting and
prioritization. The domain name used in the SRV record name is the
realm name. Several different Kerberos-related service names are
used:
_kerberos._udp
This is for contacting any KDC by UDP. This entry will be used
the most often. Normally you should list port 88 on each of your
KDCs.
_kerberos._tcp
This is for contacting any KDC by TCP. The MIT KDC by default
will not listen on any TCP ports, so unless you've changed the
configuration or you're running another KDC implementation, you
should leave this unspecified. If you do enable TCP support,
normally you should use port 88.
_kerberos-master._udp
This entry should refer to those KDCs, if any, that will
immediately see password changes to the Kerberos database. If a
user is logging in and the password appears to be incorrect, the
client will retry with the master KDC before failing with an
"incorrect password" error given.
If you have only one KDC, or for whatever reason there is no
accessible KDC that would get database changes faster than the
others, you do not need to define this entry.
_kerberos-adm._tcp
This should list port 749 on your master KDC. Support for it is
not complete at this time, but it will eventually be used by the
:ref:`kadmin(1)` program and related utilities. For now, you will
also need the **admin_server** variable in :ref:`krb5.conf(5)`.
_kpasswd._udp
This should list port 464 on your master KDC. It is used when a
user changes her password. If this entry is not defined but a
_kerberos-adm._tcp entry is defined, the client will use the
_kerberos-adm._tcp entry with the port number changed to 749.
The DNS SRV specification requires that the hostnames listed be the
canonical names, not aliases. So, for example, you might include the
following records in your (BIND-style) zone file::
$ORIGIN foobar.com.
_kerberos TXT "FOOBAR.COM"
kerberos CNAME daisy
kerberos-1 CNAME use-the-force-luke
kerberos-2 CNAME bunny-rabbit
_kerberos._udp SRV 0 0 88 daisy
SRV 0 0 88 use-the-force-luke
SRV 0 0 88 bunny-rabbit
_kerberos-master._udp SRV 0 0 88 daisy
_kerberos-adm._tcp SRV 0 0 749 daisy
_kpasswd._udp SRV 0 0 464 daisy
Clients can also be configured with the explicit location of services
using the **kdc**, **master_kdc**, **admin_server**, and
**kpasswd_server** variables in the :ref:`realms` section of
:ref:`krb5.conf(5)`. Even if some clients will be configured with
explicit server locations, providing SRV records will still benefit
unconfigured clients, and be useful for other sites.
.. _db_prop:
Database propagation
--------------------
The Kerberos database resides on the master KDC, and must be
propagated regularly (usually by a cron job) to the slave KDCs. In
deciding how frequently the propagation should happen, you will need
to balance the amount of time the propagation takes against the
maximum reasonable amount of time a user should have to wait for a
password change to take effect.
If the propagation time is longer than this maximum reasonable time
(e.g., you have a particularly large database, you have a lot of
slaves, or you experience frequent network delays), you may wish to
cut down on your propagation delay by performing the propagation in
parallel. To do this, have the master KDC propagate the database to
one set of slaves, and then have each of these slaves propagate the
database to additional slaves.
See also :ref:`incr_db_prop`
|
