Configuration
Configuration
Allowing ax25 connects from users
This is dealt with in the previous section
Setting up telnet connects (from 1.47 onwards)
From version 1.47 you can choose to allow the perl cluster.pl program to allow connections directly (i.e. not via the /spider/src/client interface program). If you are using Windows then this is the only method available of allowing incoming telnet connections.
to make the change happen...
Having done that, you need to copy the file /spider/perl/Listeners.pm to /spider/local and then edit it. You will need to uncomment the line containing "0.0.0.0" and select the correct port to listen on.
It comes out of the box looking something like:-
@listen = ( # ["0.0.0.0", 7300], );
Change it so that it looks like this:-
@listen = ( ["0.0.0.0", 7300], );
Later versions have more comments in the Listeners.pm file that are designed to help you remove the correct '#' character.
As standard, the listener will listen on all interfaces simultaneously. If you require more control than this, you can specify each interface individually:-
@listen = ( ["gb7baa.dxcluster.net", 7300], ["44.131.16.2", 6300], );
This will only be successful if the IP addresses on each interface are static. If you are using some kind of dynamic IP addressing then the 'default' method is the only one that will work.
Restart the cluster.pl program to enable the listener.
One important difference with the internal listener is that no echoing is done by the cluster program. Users will need to set 'local-echo' on in their telnet clients if it isn't set automatically (as per the standards). Needless to say this will probably only apply to Windows users.
The standard and default telnet port is 23. However, port 23 is not available to non-privileged tasks (non-root). It is NOT advisable to run spider as root. If your node does not use IPV6, set your firewall to forward port 23 to 7300 or the port you specified for spider to listen on. This will be very dependent on what your firewall and/or router configuration allows.
IP Version 6
Port forwarding is not available with IPV6. One way around the issue is to give the perl executable "low-port" privileges. On CentOS and I believe Ubuntu, there is a package called libcap. After installing the setcap utility issue the following as root on your node:-
setcap cap_net_bind_service=epi /usr/bin/perl
If the perl executable gets updated, the command will need to be issued again after the update.
The IPV6 Listeners file should look like:-
@listen = ( ["::", 7300], ["::", 23], );
This will listen on ports 23 and 7300 on ALL interfaces including any defined IPV4 and IPV6 interfaces. I have used this successfully with CentOS Version 6 and my information indicates it should work with Ubuntu as well.
Allowing telnet connects from users (before version 1.47 or for special purposes)
From version 1.47 there is a new (more efficient) way of doing this (see previous section) but, if you prefer, the method of doing it described here will continue to work just fine.
Allowing telnet connections is quite simple. Firstly you need to add a line in /etc/services to allow connections to a port number, like this ....
spdlogin 8000/tcp # spider anonymous login port
Then add a line in /etc/inetd.conf like this ....
spdlogin stream tcp nowait root /usr/sbin/tcpd /spider/src/client login telnet
Once this is done, you need to restart inetd like this ....
killall -HUP inetd
Now login as sysop and cd spider/src. You can test that spider is accepting telnet logins by issuing the following command ....
./client login telnet
You should get a login prompt and on issuing a callsign, you will be given access to the cluster. Note, you will not get a password login. There seems no good reason for a password prompt to be given so it is not asked for.
Assuming all is well, then try a telnet from your linux console ....
telnet localhost 8000
You should now get the login prompt and be able to login as before.
Setting up for AGW Engine (1.47 onwards)
AGW Engine is a Windows based ax25 stack. You can connect to an AGW engine from Linux as well as Windows based machines.
In order to enable access to an AGW Engine you need to copy /spider/perl/AGWConnect.pm to /spider/local and edit it. Specifically you must:-
- set $enable to 1.
- set $login and $passwd to the values set up in your AGW installation. If you haven't set any there, then you should not touch these values.
- You can connect to a remote AGW engine (ie on some other machine) by changing $addr and $port appropriately.
- Restart the cluster.pl program
Setting up node connects
In order to allow cluster node connections, spider needs to know that the connecting callsign is a cluster node. This is the case whether the connect is incoming or outgoing. In spider this is a simple task and can be done in runtime.
Later versions of Spider can distinguish different software and treat them differently. For example, the WCY beacon cannot be handled by AK1A type nodes as AK1A does not know what to do with PC73. There are 5 different types of node at present and although they may not have any major differences at the moment, it allows for compatibility. The 5 types are ...
set/node (AK1A type) set/arcluster set/spider set/dxnet set/clx
For now, we will assume that the cluster we are going to connect to is an AK1A type node.
Start up the cluster as you did before and login as the sysop with client. The cluster node I am wanting to make a connection to is GB7BAA but you would obviously use whatever callsign you required. At the prompt type ...
set/node gb7baa
The case does not matter as long as you have a version of DXSpider later than 1.33. Earlier versions required the callsign to be in upper case.
That is now set, it is as simple as that. To prove it, login on yet another console as sysop, cd to spider/src and issue the command ...
./client gb7baa (using the callsign you set as a node)
You should get an initialisation string from DXSpider like this ...
./client gb7baa PC38^GB7MBC^~
If the callsign you just set up as a cluster node is for an incoming connect, this is all that needs to be done. If the connection is to be outgoing then a connection script needs to be written.
Sometimes you make a mistake... Honest, it does happen. If you want to make a node back to being a normal user, regardless of what type it is, do:
unset/node gb7baa
Connection scripts
Because DXSpider operates under Linux, connections can be made using just about any protocol; AX25, NETRom, tcp/ip, ROSE etc are all possible examples. Connect scripts live in the /spider/connect directory and are simple ascii files. Writing a script for connections is therefore relatively simple.
The connect scripts consist of lines which start with the following keywords or symbols:-
- # - All lines starting with a # are ignored, as are completely blank lines.
- timeout - timeout followed by a number is the number of seconds to wait for a command to complete. If there is no timeout specified in the script then the default is 60 seconds.
- abort - abort is a regular expression containing one or more strings to look for to abort a connection. This is a perl regular expression and is executed ignoring case.
- connect - connect followed by ax25, agw (for Windows users) or telnet and some type dependent information. In the case of a telnet connection, there can be up to two parameters. The first is the ip address or hostname of the computer you wish to connect to and the second is the port number you want to use (this can be left out if it is a normal telnet session). In the case of an ax25 session then this would normally be a call to ax25_call or netrom_call as in the example above. It is your responsibility to get your node and other ax25 parameters to work before going down this route!
- ' - is the delimiting character for a word or phrase of an expect/send line in a chat type script. The words/phrases normally come in pairs, either can be empty. Each line reads input from the connection until it sees the string (or perl regular expression) contained in the left hand string. If the left hand string is empty then it doesn't read or wait for anything. The comparison is done ignoring case. When the left hand string has found what it is looking for (if it is) then the right hand string is sent to the connection. This process is repeated for every line of chat script.
- client - client starts the connection, put the arguments you would want here if you were starting the client program manually. You only need this if the script has a different name to the callsign you are trying to connect to (i.e. you have a script called other which actually connects to GB7DJK-1 [instead of a script called gb7djk-1]).
There are many possible ways to configure the script but here are three examples, one for a NETRom/AX25 connect, one for AGW engines and one for tcp/ip.
timeout 60 abort (Busy|Sorry|Fail) # don't forget to chmod 4775 netrom_call! connect ax25 /usr/sbin/netrom_call bbs gb7djk g1tlh # you can leave this out if you call the script 'gb7dxm' client gb7dxm ax25
timeout 60 abort (Busy|Sorry|Fail) # this does exactly the same as the previous example # the '1' is the AGW port number to connect thru for g1tlh connect agw 1 g1tlh # you can leave this out if you call the script 'gb7dxm' client gb7dxm ax25
timeout 15 connect telnet dirkl.tobit.co.uk # tell GB7DJK-1 that it is connected to GB7DJK # you can leave this out if you call this script 'gb7djk' client gb7djk telnet
All these examples assume that everything is set up properly at the other end. You will find other examples in the /spider/examples directory.
Starting the connection
You start the connection, from within a sysop enabled cluster login, by typing in the word 'connect' followed by a script name like this ....
G0VGS de GB7MBC 13-Dec-1998 2041Z >connect gb7djk-1 connection to GB7DJK-1 started G0VGS de GB7MBC 13-Dec-1998 2043Z >
This will start a connection using the script called gb7djk-1. You can follow the connection by watching the term or console from where you started cluster.pl. From version 1.47 onwards, you will need to set/debug connect first. You should see something like this ...
<- D G1TLH connect gb7djk-1
-> D G1TLH connection to GB7DJK-1 started
-> D G1TLH G1TLH de GB7DJK 13-Dec-1998 2046Z >
timeout set to 15
CONNECT sort: telnet command: dirkl.tobit.co.uk
CHAT "login" -> "gb7djk"
received "
Red Hat Linux release 5.1 (Manhattan)
Kernel 2.0.35 on an i586
"
received "login: "
sent "gb7djk"
CHAT "word" -> "gb7djk"
received "gb7djk"
received "Password: "
sent "gb7djk"
Connected to GB7DJK-1, starting normal protocol
<- O GB7DJK-1 telnet
-> B GB7DJK-1 0
GB7DJK-1 channel func state 0 -> init
<- D GB7DJK-1
<- D GB7DJK-1 Last login: Sun Dec 13 17:59:56 from dirk1
<- D GB7DJK-1 PC38^GB7DJK-1^~
<- D GB7DJK-1 PC18^ 1 nodes, 0 local / 1 total users Max users 0 Uptime
0 00:00^5447^~
etc
With later versions of Spider there is a set/login command for users. This tells them when a user or node logs in or out. If you do not add a line to your scripts after the final line (or before the client line which should always be last if needed) then the login/logout information will be sent to users before the login actually completes. This means if a node is unreachable, it will continue sending logins and logouts to users even though it is not actually connecting. To avoid this use the following line ...
In a script, this might look like ...
timeout 35 abort (Busy|Sorry|Fail) connect telnet mary 3000
Telnet echo
Cluster links in particular suffer greatly from the presence of telnet echo. This is caused by the telnet negotiation itself and can create at worst severe loops. At best it creates unnecessary bandwidth and large logfiles! There are things that can be done to limit this problem but will not always work dependent on the route taken to connect.
Telnet echo itself should only be a problem if the connection is being made to the telnet port (23). This port uses special rules that include echo negotiation. If the connection is to a different port, such as 7300, this negotiation does not happen and therefore no echo should be present.
Sometimes it is not possible to make a direct connection to another node and this can cause problems. There is a way of trying to suppress the telnet echo but this will not always work, unfortunately it is difficult to be more specific. Here is an example of what I mean ...
timeout 35 abort (Busy|Sorry|Fail) connect telnet mary.lancs.ac.uk
So, the first connection is made by Spider. This is fine as Spider uses the Net_Telnet script from within perl. This actually uses TCP rather than TELNET so no negotiation will be done on the first connection. Once connected to mary.lancs.ac.uk, the command is sent to suppress echo. Now a telnet is made to a cluster node that is accepting connections on port 23. The problem with this link is that the negotiation is made by the remote machine, therefore you have no control over it. The chances are that this link will create echo and there will be no way you can stop it.
Autostarting the cluster
Ok, you should now have DXSpider running nicely and allowing connects by cluster nodes or users. However, it has to be shutdown and restarted manually. It would be much easier to have it start automatically.
This is not only a way to start the cluster automatically, it also works as a watchdog, checking the sanity of DXSpider and respawning it should it crash for any reason. Before doing the following, shutdown the cluster as you did earlier.
Login as root and bring up the /etc/inittab file in your favourite editor. Add the following lines to the file near the end ...
##Start DXSpider on bootup and respawn it should it crash DX:3:respawn:/bin/su -c "/usr/bin/perl -w /spider/perl/cluster.pl" sysop >/dev/tty7
This line works fine for RedHat distributions. It is also fine for SuSE up to 7.0. From SuSE 7.1 you need to add runlevels 2 and 5 like this ...
DX:235:respawn:/bin/su -c "/usr/bin/perl -w /spider/perl/cluster.pl" sysop >/dev/tty7
The line required for Slackware distributions is slightly different. My thanks to Aurelio, PA3EZL for this information.
DX:23:respawn:/bin/su - sysop -c "/usr/bin/perl -w /spider/perl/cluster.pl" >/dev/tty7
Ubuntu (from version 6.10) and Fedora Core (from version 9) have changed from using sysvinit to the use of upstart so using /etc/inittab won't work. Instead, create a file /etc/event.d/dxspider as follows:
#DXSpider startup script # # This service maintains a getty on tty7 from the point the system is # started until it is shut down again. start on runlevel 2 start on runlevel 3 stop on runlevel 0 stop on runlevel 1 stop on runlevel 4 stop on runlevel 5 stop on runlevel 6 respawn exec /bin/su - sysop -c "/usr/bin/perl -w /spider/perl/cluster.pl" >/dev/tty7
Ubuntu made some further changes in version 9.10, so the script needs to be placed in /etc/init/dxspider.conf:
#DXSpider startup script # # This service maintains a getty on tty7 from the point the system is # started until it is shut down again. start on runlevel [2345] stop on runlevel [!2345] respawn exec /bin/su - sysop -c "/usr/bin/perl -w /spider/perl/cluster.pl" >/dev/tty7
This will automatically start DXSpider on tty7 (ALT-F7) on bootup and restart it should it crash for any reason.
- NB: It should be noted that /dev/tty7 is only an example. Some SuSE systems will only accept upto tty6. It really does not matter which tty you run it on.
As root type the command 'telinit q'. DXSpider should start up immediately. You will see the output on tty7 and if you login as sysop you should find everything running nicely.