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
218
219
220
221
222
|
General Information
-------------------
To upgrade from a previous release simply follow the
installation instructions. Your configuration files will
not be touched. New options have the older behaviour as
default, if not set explicitly. I try to keep backward
compatibility forever.
To Version 2.5 or higher
------------------------
Upgrading to 2.5 or higher requires to first upgrade the
server. It can handle requests from older clients, but
newer clients fail on an older server.
To version 3.0 or higher
------------------------
The name of the service name changed from to afbackup
from backup (important for the entry in /etc/services).
Keep this in mind, when you are going to modify the
inetd's configuration.
Check your configuration for commands, that consist of
several subcommands separated by semicolons. It was no
problem in earlier versions, but is now crucial to make
sure, that these commands exit with a bad status (!= 0),
if one of the subcommands fails. Thus the subcommands
should be separated by && instead of the semicolon. This
does not always make sense. In some cases a good exit
status (0) must be enforced. In these cases you should
write: <commands-that-might-fail>; exit 0
In some cases the server fails to run in buffering mode.
Why this happens, is currently unknown, but work is done
investigating this problem. In such a case error messages
show up for no obvious reasons. As a temporary solution
until the problem is fixed the server can be started with
the option -b. Add -b in the file /etc/inetd.conf after
the word afserver, so the whole line looks somewhat like
this:
afbackup stream tcp nowait backup /usr/backup/server/bin/afserver afserver -b /usr/backup/server/lib/backup.conf
If you have the xinetd running instead of inetd, modify
the section in /etc/xinetd.conf to be like this:
service afbackup
{
flags = REUSE NAMEINARGS
socket_type = stream
protocol = tcp
wait = no
user = backup
server = /usr/backup/server/bin/afserver
server_args = /usr/backup/server/bin/afserver -b /usr/backup/server/lib/backup.conf
}
Having successfully modified /etc/(x)inetd.conf, send a
hangup signal to the (x)inetd to activate the change. Find
out the PID typing ps -uxa|grep inetd and enter:
kill -HUP <PID>
with <PID> replaced by the real process id. The -b option
turns off buffering, so performance is worse, but until
the problem is fixed, there's no way around this, sorry.
If you make the backup into a filesystem instead of a
tape, check for the recommended settings of new parameters
in the HOWTO. The default might not be appropriate for
this configuration.
You might want to use DES encryption instead of the old
shuffeling algorithm. 128 Bit DES encryption can be used
now, what is currently and i think for the next 100 years
not crackable. Note, that whether using DES or not is a
compile time option and no runtime setting. Thus clients
and server must be compiled in the same manner. To get
the appropriate DES library, go to the ftp-directory,
where you got this package from. The current version by
the time of writing this is 4.04b.
To version 3.1 or higher
------------------------
If using several backup servers for one client, the server
and port used for this client up to now should be the first
one supplied in the client side configuration. I.e. the
new entries should be appended to the existing ones for the
backup server hostname, port number and cartridge set number
(see: HOWTO Q9).
The program name in column 7 of the /etc/inetd.conf entry
must be the full path to the program. It is no longer sufficient
to just give the basename of the program. The Install script
considers this, but if you upgrade manually, /etc/inetd.conf has
to be edited. The same applies for the first word in the
server_args parameter of the xinetd, if this daemon is used
instead of inetd.
To version 3.2 or higher from versions < 3.2
--------------------------------------------
When going to use the multi-stream server, the existing
single-stream configuration should be kept and the new
server should get a new service (TCP port) assigned. This
way existing backups don't become useless and restoring
will transparently use the old server. It is not possible
to get the data of backups written with the single-stream
server through the multi-stream server, but provided to
have the single-stream server working as configured before
this is not necessary.
Cause the emergency recovery mechanism changed heavily,
a migration must be performed on the clients. The script
migrate_to_3.2 in the distribution directory has been
written to help migrating. Started as root, it will ask
some stupid questions and then convert the old minimum
restore info to the new format. The minimum restore info
will also be written to the configured backup media doing
a small backup of the file with it. So emergency recovery
is possible also without the minimum restore info, but
better don't rely on it. Keep the minimum restore info
somewhere. It is no longer cumulative i.e. you need only
the latest one to restore everything.
The old style minimum restore information is required to
perform this step. See the configuration parameter
StartupInfoProgram on your clientside configuration,
where it is saved.
To version 3.2.2 or higher from versions < 3.2.2
------------------------------------------------
Built-in compression is available as new feature in
version 3.2.2 . When choosing to use built-in compression
see Q 17 of the HOWTO AND Q27 (!!!) of the FAQ how to
build and configure it.
Note, that nothing changed on the server side, so an
update of the server will not have any effect.
To version 3.2.2 or higher from versions < 3.2.3
------------------------------------------------
The server side configuration file entry to specify the
cartridge sets has been reworked. It is no longer called
LastCartridges: followed by the the numbers of the last
cartridges in the sets. Now it reads as CartridgeSets:
followed by a list of cartridge set specifiers separated
by whitespace. See CONFIG or the manual pages for valid
formats.
The server program understands both versions, but if
both are present, the old style entry is ignored. Also
the configuration programs convert the configuration file
entry to the new format silently. The recommended way to
switch to the new style entry is nonetheless to run the
script migrate_to_3.2.3 . It asks for the full path to
the configuration file, makes a safety copy of it and
converts it to the new format. The differences between
old and new file are printed, so you can see, what has
changed and if the automatic conversion fails, have a
chance to adapt the file manually.
To version 3.3 or higher from versions < 3.3
--------------------------------------------
Changes in the encryption routines were necessary due to
a bug in all libdes versions. Thus the newly built programs
are unable to authenticate successfully with existing
installations. Furthermore if the __descrpt program was used
to encrypt files, they cannot be decrypted with this program
any longer. Nonetheless it is possible to build the package
using the buggy function, so backward compatibility for
existing installations is provided. If you are using the
Install script, answer the question for backward compatibility
appropriately. If you prefer to install step by step, don't
miss the section about this in the INSTALL file.
To version 3.3.4 or higher from versions < 3.3.4
------------------------------------------------
If it is desired to downgrade to earlier versions because of
problems with this one, the serverside tapepos file must be
converted back to the old format. The new format is e.g.:
/dev/nst0 9 1
1: 19 14
2: 9 10
3: 11 1
This means: In device /dev/nst0 there is cartridge 9 wound back
to tapefile 1. The next 3 lines indicate the writing position
for 3 cartridge sets. To convert back to older format, take
only the 2 trailing numbers from each line and put them into
1 single line, beginning with the line starting with the
device specifier, then following up the cartridge set numbers
in serial order. This would for this example result in the
following line:
9 1 19 14 9 10 11 1
To version 3.3.6 or higher from versions < 3.3.6
------------------------------------------------
The release of 3.3.6 was originally not planned. Modifications
for the upcoming 4.0 were already done and i didn't feel like
rolling them all back. So if upgrade to 3.3.6 has been done,
but for some reason a downgrade to a lower version is desired,
the configuration files must be moved back from the new .../etc
directories to the appropriate .../lib .
|
