2 Ian Maude, G0VGS, (g0vgs@gb7mbc.net)
3 February 2005 Revision: 1.8
5 A reference for SysOps of the DXSpider DXCluster program.
6 ______________________________________________________________________
12 1.1 Where do I get DXSpider?
13 1.2 How do I use the patches?
14 1.3 If I use a tarball to overwrite my installation, what happens to my configuration?
15 1.4 I am running RedHat 5.2 and I am getting strange errors, what is wrong?
18 2.1 How can I get Spider to restart automatically if it crashes?
19 2.2 How can I monitor traffic to and from a node or user?
20 2.3 I see spots coming in my debug log, but none go out to the users
21 2.4 My neighbouring node cannot use the RCMD command to me, he just keeps getting the "tut tut" message.
22 2.5 I do not seem to be sending any bulletin mail to my link partners, what is wrong?
23 2.6 How can I automatically limit the amount of debug logfiles that are stored?
24 2.7 I updated my Linux distribution and now Spider cannot read the users file or the dupefile, what is the problem?
25 2.8 Since I last updated I seem to be getting duplicate spots appearing.
26 2.9 I have deleted a message but it is still there, why?
27 2.10 I am updating from CVS and I get all sorts of errors when I restart
28 2.11 I have done a CVS update, restarted and it says that "fileX" is missing
31 ______________________________________________________________________
33 Please mail any FAQs to the maintainer at the address above.
37 \e[1m1. Installation
\e[0m
39 \e[1m1.1. Where do I get DXSpider?
\e[0m
41 All things Spider can be found at www.dxcluster.org and the actual
42 program and patches can be found in the download area.
45 \e[1m1.2. How do I use the patches?
\e[0m
47 Patching is done in the standard linux way ...
51 zcat /tmp/patch-1.40 | patch -p0
55 assuming the patch version you are using is 1.40 and resides in /tmp!
58 Be aware that each patch assumes the previous patch has been applied.
59 That is to say if you are patching from version 1.38 to 1.40 you would
60 first need to apply patch-1.39 and then patch-1.40.
64 \e[1m1.3. If I use a tarball to overwrite my installation, what happens to
\e[0m
65 \e[1mmy configuration?
\e[0m
67 The tarballs are designed to not overwrite your existing configuration
68 and can be used at any time to update your Spider software. All the
69 key files have the suffix .issue (eg. DXVars.pm.issue) at default.
72 \e[1m1.4. I am running RedHat 5.2 and I am getting strange errors, what is
\e[0m
75 The version of Perl that comes with 5.2 seems to be some kind of pre-
76 release and is broken. You can get a new version of perl from
77 www.dxcluster.org or ftp://updates.redhat.com. Don't forget to patch
78 it with the CPAN modules.
81 \e[1m2. Administration
\e[0m
83 \e[1m2.1. How can I get Spider to restart automatically if it crashes?
\e[0m
85 Put this line into /etc/inittab ..
89 DX:3:respawn:/bin/su -c "/usr/bin/perl -w /spider/perl/cluster.pl" sysop > /dev/tty7
93 Run
\e[4mtelinit
\e[24m
\e[4mq
\e[24m as root. Spider will restart so be aware. However, any
94 time you reboot, cluster.pl will start in tty7 and if it crashes, it
98 \e[1m2.2. How can I monitor traffic to and from a node or user?
\e[0m
100 There are 2 ways to achieve this. You can use the
\e[4mtail
\e[24m command like
105 tail -f /spider/data/debug/167.dat |grep G0VGS
109 or in later versions of Spider, there is a command called
\e[4mwatchdbg
\e[24m in
110 which case you simply type ..
118 \e[1m2.3. I see spots coming in my debug log, but none go out to the users
\e[0m
120 Please check the time on your PC.
123 All spots are checked that they are no more than 15 minutes in the
124 future and 60 minutes in the past. If your clock is not set to GMT
125 (UTC) and is more than one hour out (say on your local (summer) time)
126 then the test will fail and no spots will come out. Neither will they
130 If you are connected to the internet, most linux distributions have an
131 implementation of ntpd. The Microsoft Windows 2003, XP, 2000 and NT
132 machine clock can also be set to be syncronised to an NTP source.
133 This can be done in the standard time configuration screen. There is
134 also the simple nettime program for Windows 95/98/ME.
138 \e[1m2.4. My neighbouring node cannot use the RCMD command to me, he just
\e[0m
139 \e[1mkeeps getting the "tut tut" message.
\e[0m
141 Assuming that the permissions are set correctly (perm level 5
142 required), it could be that the home_node is set incorrectly. You can
143 reset the home_node using the
\e[4mspoof
\e[24m command like this ..
147 spoof gb7adx set/home gb7adx
151 Assuming that the node_call you are changing is gb7adx.
154 \e[1m2.5. I do not seem to be sending any bulletin mail to my link part-
\e[0m
155 \e[1mners, what is wrong?
\e[0m
157 There is a file in /spider/msg called forward.pl.issue. Rename this
158 to forward.pl and edit it to meet your requirements. You will need to
159 issue the command
\e[4mload/forward
\e[24m or restart Spider for the changes to
163 \e[1m2.6. How can I automatically limit the amount of debug logfiles that
\e[0m
164 \e[1mare stored?
\e[0m
166 Use the tmpwatch command. Create a file in /etc/cron.daily/
167 containing the line ...
171 /usr/sbin/tmpwatch -f 240 /spider/data/debug
175 Remember to make it executable!
177 This will limit your debug data down to the last 10 days
180 \e[1m2.7. I updated my Linux distribution and now Spider cannot read the
\e[0m
181 \e[1musers file or the dupefile, what is the problem?
\e[0m
183 Almost certainly this is a change in the db format of perl. Follow
184 these few steps to correct the problem.
187 o stop the cluster (disable any autostart in inittab)
192 o issue the command perl user_asc
194 o restart the cluster
196 That should solve the problem.
199 \e[1m2.8. Since I last updated I seem to be getting duplicate spots
\e[0m
202 What has probably happened is that the dupefile has got corrupted in
203 some way. Simply delete the /spider/data/dupefile and restart the
204 cluster. It may take a little time to become fully functional but
205 should solve your problem.
208 \e[1m2.9. I have deleted a message but it is still there, why?
\e[0m
210 This is now the way messages are handled for deletion in Spider. If
211 you look closely you will see a 'D' following the message number.
212 This message is marked for deletion and will be deleted in 2 days if
213 nothing further is done. Optionally you can use the command
214 \e[4mdelete/expunge
\e[24m to delete it immediately.
217 \e[1m2.10. I am updating from CVS and I get all sorts of errors when I
\e[0m
220 Firstly, whenever you update from CVS, a log is displayed. Next to
221 each file that is downloaded there is a letter, e.g.:
230 cvs server: Updating perl
241 For normal CVS use you should only ever see the letter 'P'. This means
242 that the file has changed in CVS and is more recent than the one that
243 is currenly on your system. You may also possibly see the letter '?',
244 which means that there is a file that you (or the system) has created
245 that CVS doesn't know about and isn't under its control. These are all
249 Sometimes you will see the letter 'U' next to a file. This means that
250 it is a new file that you don't currently have. This is also OK.
254 However, if you see the letter 'C' or 'M', that means that CVS thinks
255 that the file has changed locally. For the letter 'C', it has changed
256 sufficiently near to one or more modifications which CVS wants to
257 download to your system. For the 'M', CVS thinks that it is OK to make
258 the change (you may also see some messages about "merging revision
259 1.xx with 1.yy"). Neither of these things are good. Files that are
260 under the control of CVS must not be changed by sysops. It is the
261 files that have the 'C' next to them that will show the errors and
262 they will be things like:-
266 Syntax error near '<<<<' at line 23
267 Syntax error near '===' at line 40
268 Syntax error near '>>>' at line 51
272 You will now necessarily see all of the errors shown but you will get
273 one or more sets of some of them. The cure is simple:
276 o identify the file that is causing the problem.
280 o run the cvs update again.
283 You will see that file come back (with a letter 'U' next to it). That
284 will be the correct file as CVS thinks it should be. If you
\e[4mstill
\e[24m have
285 a problem, then get onto the dxspider-support mailing list.
288 If all else fails (or you have several conflicts) you can safely
289 remove the entire /spider/perl and /spider/cmd directories and then
290 run the cvs update. They will all be recreated in their pristine
294 \e[1m2.11. I have done a CVS update, restarted and it says that "fileX" is
\e[0m
297 The correct way to run cvs is:-
306 The '-d' is crucial. It makes sure that any new directories that may
307 contain new code are created and that new code downloaded. I have
308 absolutely no idea why this is not the default, seeing as CVS (in
309 other circumstances) happily recurses its merry way down a directory
310 tree, but there you are.
313 WinCVS and other graphical CVS frontends have a checkbox for the
314 update screen called something like "create sub-directories" (it may
315 be hidden in some sub-screen - go look for it if it isn't obvious).
316 Make sure that this box is checked. If you can make this the default
317 setting in the program's setup screen then please do that. It will
318 save you a lot of pulled hair.