home *** CD-ROM | disk | FTP | other *** search
Text File | 1993-08-23 | 51.6 KB | 1,365 lines
Last Changed: $Id: FAQ-inn.3,v 1.39 1993/08/19 17:42:42 tal Exp $ The INN CONFIGURATION FAQ ------------------------- by Tom Limoncelli, with additions by many, many others. (Copyright 1992,1993 Tom Limoncelli, this may be redistributed in whole and only on electronic networks. Otherwise, contact the author. Publishers and movie agents should contact me directly. I'd prefer it if Tom Cruise plays me in the movie, but if he isn't available we can negotiate.) A special thanks goes to all the people that have helped put this together. I've tried to give credit where credit is due. Additions and corrections should be sent to tal@Warren.MENTORG.COM. DO NOT send requests for help to Tom. Post them to news.software.nntp. Table Of Contents: ------------------ GENERAL OVERVIEW OF INN Should I read the install.ms file in its entirety before reading this document? How does it all fit together? (LONG) Terminology used in the rest of this document. What should I monitor as I debug INN problems. CONFIGURATION DEBUGGING GUIDE Connecting to a TCP/IP server. My innd won't start! Make sure that "feeders" can connect. Make sure that "readers" can connect. Make sure that clients can post. "client" doesn't have the software needed to post. Debugging the "newsfeeds" file. The ME line in the newsfeeds file. Cookbook example of an outgoing NNTP feed: Cookbook example of an outgoing UUCP feed: Testing an outgoing feed (your "newsfeeds" configuration). Cookbook example setting up NOV ("overchan"). STARTUP PROBLEMS AND SYSLOG MESSAGES syslog message: ME internal no to group syslog message: ME cant sendto CCreader bytes 4 No such file or directory. syslog message: ME internal no control and/or junk group syslog message: ME bad_newsfeeds no feeding sites Why do all these "readclose" messages show up in my syslog? "File exists writing symlinking article file -- throttling" "cant fopen <newsgroup>/.thread No such file or directory" MISC QUESTIONS Can I edit my configuration files where they are, or do I have to edit them in $INN/site ? syslog message: ME cant nonblock 15 Operation not supported. Suddenly my active and history files are owned by root! What can I do if I can't purchase the O'Reiley And Associates book on Managing Usenet? Getting an active file. How do I use nntplink with INN? How do I use newsgate with INN? Other cron jobs. After a crash. Debugging someone that is feeding you. Feeds suddenly can't connect anymore! How does the "ME" line interact with the other lines? Why doesn't nntpget work? More about the "to.*" groups. I'm getting groups sent to me that I don't want. When my feeder connects, I get articles but they don't take what's waiting for them. Directories are being created with wrong permissions. Can I split my news spool across partitions? Warnings to people that must set HAVE_UNIX_DOMAIN to DONT NOTE: Read this document through from beginning to end after you first compile INN. Later, use it for reference if problems come up. ====================================================================== GENERAL OVERVIEW OF INN ====================================================================== ------------------------------ Subject: Should I read the install.ms file in its entirety before reading this document? YES! Install.ms tells you how to compile and install the software. This document walks you through debugging the *configuration* of the software once it is installed. This document takes you from where install.ms leaves off, gives you a quick overview of how all the pieces fit together, and then takes you through specific debugging tasks. Debugging INN problems is often difficult because one needs to be an experienced netnews person to do it well. You can only get experience by having a properly running system. This is a catch-22. This tutorial attempts to take you through the basics. The rest you'll figure out. Newsgroups you should know exist: news.software.nntp -- INN questions go here. news.software.b -- Discussions about any of the many software packages that support the "B news" format (i.e. INN, C news, ANU-NEWS, etc.) This document also takes you through the process of verifying that your system is properly configured. When you are done, you should: 1. be sure that when feeders connect they are treated as feeders. 2. be sure that when clients connect they are treated as clients. 3. be sure that posting works. 4. be sure that your out-bound feeds are properly configured. ------------------------------ Subject: How does it all fit together? (LONG) Here is a fantastic overview of the workings of INN. From: Ken Hornstein <kenh@leps5.phys.psu.edu> I discovered that the biggest problem I had with INN was understanding how everything fits together (since I had no experience with B or C news). Here's a (hopefully) simple description of how everything fits together: After running rc.news (as "root"), you should have the "innd" daemon running ("ps" will show the process to be owned by "news"). This is the Master Daemon. It handles incoming connections, stores the articles on your disk, but does _not_ send any articles out itself. It directs other programs to do that. Exactly where articles are sent and how they are sent is determined by the "newsfeeds" file. Setting up your newsfeeds file will be the hardest part of configuring INN. Here are some example entries from my newsfeeds file: ra/ra.nrl.navy.mil\ :*,!psu.*/!psu\ :Tf,Wnm: Looks complicated? It isn't. Here's what it means: "ra" is the name of the feed. "/ra.nrl.navy.mil" is an alias for ra. This is important because INN uses the "Path" header to insure the articles are not sent to sites where they have already been. Thus, any article that has "ra" or "ra.nrl.navy.mil" in the Path header will NOT be sent to this site. The second line tells what articles will be sent to this site. "*,!psu.*" means that all articles that are not in psu.* will be sent to ra. The details of the pattern matching is found in the wildmat(3) man page. The "/!psu" means that articles with a "Distribution" header of psu will also not be sent to ra. The last field specifies exactly what _kind_ of feeds. "Tf" means this is a file feed. Unless you have unusual requirements, all of your feeds will be file feeds. "Wnm" means that the relative path name and the Message-ID of the article will be written to this file. By default, this file is called the same name as your feed file, and is in your out.going directory. So on my system, every article destined to ra will have its filename and Message-ID written to the file "/var/spool/news/out.going/ra". So how do the articles actually GET to ra? You run a program that reads the feeds file and transmits the article. Two such programs are included with INN -- "send-nntp" and "nntpsend". My personal preference is for nntpsend. If you are going to use nntpsend, you will need to add a similar line to your nntpsend.ctl file: ra:ra.nrl.navy.mil This tells nntpsend that articles in the feed file "ra" should be sent to the site "ra.nrl.navy.mil". I run nntpsend out of cron every 10 minutes with this line (in /usr/lib/crontab): 0,10,20,30,40,50 * * * * /bin/su news -c '/usr/local/news/bin/nntpsend' This is under Ultrix. A sane cron would let you specify the userid to run programs under. UUCP feeds work similarly and are described in a different section. As each article comes in (note that hosts feeding you _must_ be listed in the hosts.nntp file), innd will examine it and distribute to your listed feeds based on the above-described selection criteria. One last important thing to do is to make sure your articles get expired. This is done from the "news.daily" script. The "expire.ctl" file describes how long you want each article to last. Here are some sample lines from my expire.ctl: /remember/:14 This line tells expire to keep history entries for articles 14 days after they have been deleted. *:A:1:7:21 This is the default line. This says that by default, an article is kept a minimum of one day, the default expiration time is 7 days (this applies if there is no "Expires" header), and the very maximum that the article is kept is 21 days. psu.*:A:1:14:28 This line applies to groups only in Penn State. By default, those articles will last 14 days, 28 days at the most. Note that lines in expire.ctl should have the most general entries first, with the most specific entries last. ------------------------------ Subject: Terminology used in the rest of this document. We will pretend that your machine is named "nntphost" or "nntphost.do.main" and that there is a client named "client" or "client.do.main". Some machines connect to you to try to feed you new articles. We'll call these machines "feeders". Some machines try to connect to you to read and/or post articles. We'll call these machines "readers". ------------------------------ Subject: What should I monitor as I debug INN problems. 1. run "tail -f /var/adm/messages" to see if any syslog messages are being generated. 2. run "tail -f /var/log/news/news.err" to see if any fatal errors happen. 3. Check for incoming email constantly (especially when trying to post from "nn"). ====================================================================== CONFIGURATION DEBUGGING GUIDE ====================================================================== ------------------------------ Subject: Connecting to a TCP/IP server. You know that "telnet"'ing to a machine lets you log into it. Many TCP/IP services allow you to "telnet" into their port and talk directly to them. Try "telnet nntphost 21". This means log into port #21 (the "ftp" port) instead of the usual remote login port. Once you are in, you'll get no prompt. Type "help" and press RETURN. You should get a list of commands. If you know what the commands are, you can talk to this server. Type "quit" and press RETURN to get out. After every command you should get some kind of status message. Each line will begin with a number. Each message has a unique number. Errors are defined as anything that starts with a number >= 400. SMTP (mail) and NNTP (netnews) work the same way. Telnet into their port and issue commands and data. "quit" always gets you out. We'll use this to debug INN configurations by "telnet"'ing into the innd server and seeing the raw error messages it gives us. Try "telnet"'ing into the NNTP port (#119) of a working NNTP server to see what it's like. ------------------------------ Subject: My innd won't start! Keep a "tail -f /var/adm/messages" running. INN reports most errors via syslog. The syslog messages usually explain what is wrong. Elsewhere in this document are details about some of the less obvious syslog messages. Chances are, INN is starting, finding a misconfigured "ME" line in the newsfeeds file, and exiting. You might want to read the section on configuring your "newsfeeds" file first. Rich Salz says a common reason is that you ran makehistory but didn't rename the DBZ files. Izar Tarandach <izar@cs.huji.ac.il> suggests that another common mistake is that innd wasn't being started by the correct uid. innd (and therefore rc.news) must be started from "root" (not "news"). It immediately turns itself in user "news" once certain tasks are completed. ------------------------------ Subject: Make sure that "feeders" can connect. "feeders" are listed in hosts.nntp. "readers" are listed in nnrp.access. This section deals with "feeders" and hosts.nntp. When a machine connects to the NNTP port of nntphost, it connects to the innd process. innd knows the internet address of the machine that is making this connection, and sees if it matches the internet addresses many of the machines listed in the hosts.nntp file. If the machine is not listed in hosts.nntp, it is assumed that this machine is not a "feeder" and forks off a nnrpd to handle this connection as a "reader". If you didn't know that, you didn't read enough of the INN installation documentation. Go back and read it now. Read the man page hosts.nntp to get a complete understanding of what's going on. nnrpd uses its own authentication scheme, which is described in the next section. Since I know you didn't read that man page, I'll give you one more chance to read them now. Let's configure hosts.nntp. Just enter the names of all the machines that feed you: feeder1.do.main: feeder2.do.main: I don't use passwords yet. If you do, add them after the ":". Now let's test to see if the feeder can connect properly. Log into to the feeder and "telnet nntphost 119". If you can't log into a feeder, configure your own machine as a feeder (i.e. feeder to itself) for testing purposes. Once you can see that INN is treating that machine as a feeder you can replace the machine's name with the name of a real feed. If you are given an error message and booted out, check the error message to see what's wrong. Maybe the machine is running maintenance at the time and you have to try again later. Maybe the machine doesn't recognize you at all and you have to edit "hosts.nntp" (and don't forget the "ctlinnd reload" command!). If your "history" file or other files have the wrong ownership or protections INN will mention the offending file in the error message. Another common mistake is that people try to use wildcards in hosts.nntp (which is not supported). Remember, there are very few machines that you consider to be "feeders", so you don't want to use a wildcard. To test a "feeder": If "feeder1" can send an "ihave" command and get a "335" as a response, you know that "nntphost" is permitting "feeder1" to transfer news as a "feeder". "ihave" requires an operand. I usually type "ihave <1@test>" ("<1@test>" is a Message-ID that I know doesn't exist) and press RETURN. If I get "500 What?" I know that innd assumed that I'm a "reader" (so I have to edit my "hosts.nntp" file and add this client). If I get "335" and then a blank prompt, then INN is expecting to be fed an article. I usually just "^]" (control-]) and "quit" out; I know that it was willing to accept the article. If I get some other error message, it usually gives me enough information to debug the problem. ------------------------------ Subject: Make sure that "readers" can connect. As I wrote before, if a connection comes from a machine that isn't listed in the hosts.nntp file, it is assumed to be a "reader". A "feeder" can also issue the "mode reader" command to become a "reader". If you have "telnet"'ed in as a "feeder", try issuing this command. Note: If a site is going to feed *and* read, you'll have to link readers with innd's client library. The reason for this is that the clients must tell innd that they want to read using the "mode reader" command. The library does that automagically. It is rare that you have a machine that is a reader and a feeder (since people will want to read on their local machine, not yours.) News readers are now being packaged as "INN ready" so this will be less and less of a problem. Once the connection has been handed off to nnrpd, nnrpd checks to make sure you are authorized. It does that by reading the nnrp.access file. There is a problem with what you enter in that file. Namely, I might call the client machine "client", but that doesn't matter. What matters is what "nntphost" thinks "client" is called. Maybe "nntphost" thinks its name is "client.do.main" or even "137.202.177.3". It doesn't matter what *you* call "client", permissions in the nnrp.access file have to be specified based on what "nntphost" calls "client". Technically, nntpd uses gethostbyaddr() to reverse-lookup the name. gethostbyaddr() uses DNS or, if you are on a brain-dead Sun running Sun's NIS/DNS hack, it uses NIS, or DNS, or whatever the hell Sun was thinking when they created that cruft. To find out what "nntphost" thinks your machine is called, do the following: Telnet from "client" to "nntphost" and execute the "finger" command (just "finger" alone on the command line). The last column is what "nntphost" thinks your machine is called. If you don't have an account on both machines things are more difficult, consult your NIS or DNS expert to tell you what the answer would be. There is one exception to this technique. If you are using SunOS and braindead NIS you get just the machine name (like "milk") instead of the FQDN (like "milk.warren.mentorg.com") then you must tack on a period then the domain of the machine. So, with this knowledge (what your nntphost think's client's name is) and a copy of the man page, edit nnrp.access and add "nntphost"'s name for "client" to the file. Only nnrp.access can have wildcards (for example, "*.sjc.mentorg.com"). You'll want to include wildcards for all the domains that should be allowed to read/post. Here are some decent examples from my nnrp.access file: -------------------------------------- Tom's nnrp.access file START ## Default is no access, no way to authentication, and no groups. *:: -no- : -no- :!* *.mentorg.com:Read:::* *.mentor.com:Read:::* *.warren.mentorg.com:Read Post:::* -------------------------------------- Tom's nnrp.access file END The second field of "nnrp.access" is case sensitive. "read post" does not mean the same as "Read Post". If you know this already it's because you read the man page. Note: nnrpd will append the domain to a name that is not a FQDN. There is no need to try to find a wildcard that will match non-FQDN names (i.e. machines in your local NIS cluster). Previously this FAQ had reported that "*[^.]*" would match these names but that was wrong (the wildcard matches everything, oi!). nnrpd turns non-FQDN's into FQDNs. After you change "nnrp.access" you don't have do "ctlinnd reload" since the file is read by each nnrpd as they start up. Now "nntphost" should be letting "client" read. Let's test this out: Log into to the reader and "telnet nntphost 119". To test a "reader": Give the "mode reader" command and see how it likes it. If it doesn't give an error, then nnrp.access is letting you through. To read an article (or just get the header) type "head <2@test>" and press RETURN. Again, "<2@test>" is a message-id that I know doesn't exist. If you are allowed to read at all, it will tell you that it can't find that article. You might try the command with an message-id that you know exists. You are done with that phase. Now you should test to see if posting is permitted. Skip ahead to the next section. If "mode reader" gives an error (and rudely disconnects you) then you have a typo in nnrp.access OR you didn't issue the "ctlinnd reload" command correctly (or at all) OR nntphost thinks that "client" is called yet something else OR innd can't exec nnrpd for one reason or another -- see the syslog output or the innd.err log file. Go to the beginning of this section and start over. Note: Some telnet implementations are Real Stupid and disconnect you before showing the error message. You can also run nnrpd by hand if you have stdin:Read Post:::* in your nnrp.access file. Just run nnrpd and type interactively. This is useful for making sure it's compiled right. ------------------------------ Subject: Make sure that clients can post. The "inews" command (usually in /usr/local/bin) takes a post from a user, adds any missing headers, appends the first 4 lines of ~/.signature (if it exists), and possibly replaces any headers that are seriously forged. "inews" will also reject a message if you really botch it. "inews -h" expects a post on stdin beginning with headers, then a blank line, then the body. "inews -h -D" doesn't post the message, but outputs what it would have posted. The minimum headers one can feed is "Newsgroups:" (which is plural) and "Subject:" (which is singular). By the way, after a header there is exactly one colon then exactly one space. The space is a space, not a tab. Also, the list of newsgroups on the "Newsgroups:" line is a comma separated list, with no spaces. There are no spaces before the colon. If there is nothing after the colon or if there is only whitespace after the colon then that header will be removed by "inews". Sites that don't remove such "empty" headers have broken software. Get it? Got it? Good. Here's the test message I constantly use: ------------------------ cut here 8< inews -h -D Newsgroups: foo.test Subject: test of inn posting this is a test ------------------------ cut here 8< Exciting huh? You might also use the 'feedone' program in the frontends directory. Do "cd $inn/frontends ; make feedone" to get it built. To run it, do feedone -t -r </tmp/inews.output This will (-t) trace all I/O with the server and (-r) use a random message-id each time. If you want to test posting from a newsreading host (i.e., one that connects to nnrpd and uses the POST command) use the -p flag. If inews was able to get to the /usr/lib/news/inn.conf file (for defaults) you should get a nice post on your screen. If you don't, here are my suggestions: 1 -- You have an old inews from C news or B news laying around 2 -- inews will give you an error message saying what's wrong. You might want to look around the usual places to make sure that there are no old versions of "relaynews" or "inews". People trying to use the "inews" from C news will get a message about "can't open redirection" or similar. Make sure they are running the programs included with INN. There is something called "mini-inews" which should just take a post and send it to the nntp server. Delete that and replace it with INN's inews. INN's inews is mini-inews and regular inews, it is the ying and then yang of inewses. It is the one inews to end all inewses and all others are false idols. Note: False idol worshipper and heathen David Myers <dem@meaddata.com> reports that mini-inews works fine. He stays with mini-inews... "because INN inews needs to access not only inn.conf, but moderators, too. Installing and maintaining these files in a ~1000 client, multiple administrative domain setup like ours is too much of a pain. Most (all?) of the work done by INN inews is done by in.nnrpd during posting, anyway." INN's inews sometimes prints the error: "Can't get list of newsgroups, No such file or directory.". inews called CAlistactive() to get a local copy of the active file. If it can't reach the active file you get this error. Look at your PATH_TEMPACTIVE and see if it makes sense; i.e., if it is a valid /tmp directory. "inews -h" sometimes reports: Can't send article to the server: 441 480 Transfer permission denied This means that you set HAVE_UNIX_DOMAIN to DONT and you don't have your news server in its own hosts.nntp file. (nnrpd gets a POST, connects to innd over a TCP socket and sends an IHAVE.) (thanks to Chris Jackson <cjj@sun.com> for pointing this out). Add your news server's name to hosts.nntp and do "ctlinnd reload hosts.nntp". (for the reason why, read "Warnings to people that must set HAVE_UNIX_DOMAIN to DONT") If it still doesn't work, look through your syslog to see the name of the host that innd got, and why it handed off to nnrpd. Perhaps there is a DNS/NIS/hosts-file mismatch. (suggested by Rich Salz) Other problems are usually the result of not being able to find the "inn.conf" file (copy it to the client or make it available via NFS) or you are using Sun's brain-dead NIS/DNS stuff which doesn't do reverse name lookups well. If inews tells you that it can't generate a Message-ID, this is because it can't figure out your domain. Force it to know your domain by adding a "domain:" line in "inn.conf". Once you get "inews -h -D" working, do the same test without the "-D" option and let it actually post the message. If it can't post, it will tell you why. If the answer isn't clear enough, "telnet nntphost 119", give the "mode reader" command, then the "post" command. Enter lines of text like you would to "inews -h" and then type "." on a line by itself (and press RETURN). If posting via "telnet nntphost 119" DOES work and posting via "inews -h" DOES NOT work, you know that (1) "inews" is compiled wrong, or more likely, (2) you aren't using INN's inews. Either way, if this is happening you know you have narrowed your problems down to the inews program. By the way, posting to misc.test is pretty useless considering that the entire world doesn't need to see your message. Post to a local newsgroup or even a state-wide newsgroup like "nj.test" (assuming you are in New Jersey). There are lots of people that reply to every test message they see, so expect to get tons of stupid email. (though, if you don't get any email consider yourself lucky). Also, there is no newsgroup called "news.test" so don't post there. Many do, many fail. By the way, if you are one of those people that reply to every test message they see, get a real hobby. Do *NOT* post your test message to a non-test newsgroup. You will get many angry replies from all over the world. Look at the posted message in the news spool (if you post a message to nj.test, "cd /var/spool/news/nj/test" and cat the highest numbered file you see). If your site name is listed multiple times in the "Path:" header, put your server's name on the "pathhost:" line of "inn.conf" and recompile INN with "INEWS_PATH" set to "DONT". (I don't know why Rich likes that as the default!) REMEMBER: inn.conf is read into innd only once. After it is changed, the innd daemon must be shutdown and restarted. (use "ctlinnd shutdown x" and then run rc.news as root). If "inews -h" posts a message, smile because most of the battle is over. ------------------------------ Subject: "client" doesn't have the software needed to post. If the client doesn't have "inews" at all, check the INN installation manual to find out how to compile it. There is a special gimick included with INN to compile inews for the various other OS's and versions of Unix without having to compile the entire INN package. Since nnpost, Pnews, postnews, and all other news posting software shouldn't do anything but ask for header information, let you add a body, and then pipe the whole thing to "inews -h", you can be pretty certain that if "inews -h" works, your news posting programs will work. Think again! Post from each of them and make sure they all get posted. You might find that they access a copy of "inews" that was part of C news, mini-inews, or heavens knows what. I highly recommend that people use "find" or "gnufind" to seek out and replace any old versions of "inews". gnufind / /usr /usr/local /usr/lib -xdev -name inews\* -print For every one that you find, do the following: mv inews inews.cnews ln -s /usr/local/bin/inews inews Now you only have to update /usr/local/bin/inews, rather than chasing may copies. "nn" and "nnpost" create a file called "~/.nn/params" right before you post with tons of useful information. While posting you can shell out of the editor and view the file. The file is deleted after the message is posted. I had to view this file while shelled out of my editor to find which "inews" was being used by "nnpost". It's also a good idea to check your mail now and then while you are doing this. Some newsreaders (like "nn" notify you of a posting problem via mail. On non-INN systems, "inews" returns pretty quickly. Actually they fork a process to do the actual posting in the background. When those "inews" return, you don't know if the post was successful or not. These "inews"'s have a "-W" option which turns off this forking feature (i.e. Wait for the post to complete). INN's "inews" never forks because the wait is never that long. When "inews" returns you know if the post was successful or not. INN's "inews" accepts the "-W" option for compatibility. This may seem obvious, but when posting a test message, consider including the machine you are posting from and the program you are using. Even though you may check to see if the message got posted after every test, this will help you later when you go back to see what you have done. ------------------------------ Subject: Debugging the "newsfeeds" file. Outgoing news is controlled by the "newsfeeds" file. The INN 1.2 man page for this file is a bit complex. The man page in 1.3 (and beyond) gives better examples. Here's a "cookbook" of examples that should cover most of your needs. Debugging tips are also included. Always remember that newsfeeds uses "wildmat" matches, not the semi-regular expressions that C news uses. This means that if you want to get comp.foo and the subgroups under it (comp.foo.bar, comp.foo.baz, etc.) you have to use a statement like: comp.foo,comp.foo.* OR comp.foo* BUT NOT comp.foo.* ------------------------------ Subject: The ME line in the newsfeeds file. The "ME" entry is a bit confusing. Be careful when you read the man page. Here is the "ME" line that I use in my "newsfeeds" file. I find it works quite well, but you might want to remove the distributions that you don't need (i.e. New Jersey). Since my site has clients reading from all over the world I try to have every distribution I can find. However, I hear of a new distribution almost daily so this list is always changing. ME:!*/\ news,gnu,comp,biz,alt,rec,misc,sci,soc,talk,inet,world,worldwide,all,\ aus,su,uk,york,eunet,na,can,qc,tor,us,usa,mn,oh,chi,ca,ba,tx,pnw,il,ne,\ ny,nyc,phl,bl,nj,warren:: If you want to blindly accept all distributions, try this: ME:!*:: ------------------------------ Subject: Cookbook example of an outgoing NNTP feed: This example involves a machine named oddball.mentorg.com, that has an alias of oddball.sjc.mentorg.com, which should receive all posts (but control & junk should never be passed on) and not certain distributions. Add the following line to newsfeeds: oddball.mentorg.com/oddball.sjc.mentorg.com:*,!control*,!junk/!local,!warren:Tf,Wnm: Have the user "news" run the following via cron: 3,23,43 * * * * /usr/lib/news/bin/nntpsend >/dev/null 2>&1 (this only needs to be added once. nntpsend refers to a file called nntpsend.ctl to find out what to do). Add the following to nntpsend.ctl: oddball.mentorg.com:oddball.mentorg.com:: Done! ------------------------------ Subject: Cookbook example of an outgoing UUCP feed: Example: A site named "plts" that can not get the "clari" newsgroups or distribution "warren". Add the following to the newsfeeds file: plts:*,!clari.*,!junk*,!control*/!warren:Tf,Wnb: Add the following to the cron tab (as user "news"): 0 0-5,16-23 * * 1-5 /usr/lib/news/bin/sendbatch -c plts >/dev/null 2>&1 ------------------------------ Subject: Testing an outgoing feed (your "newsfeeds" configuration). Here is a decent game-plan for testing your newsfeeds configuration: Suppose your site is in New Jersey and you have a distribution called "mentorg" which should be used by people that want to make sure that their post will not leave their company (Mentor Graphics). You should do a test post to "nj.test" with no "Distribution:" header, and with "Distribution: nj" and "Distribution: mentorg". After posting, do a "ctlinnd flush ''" and make sure that the /var/spool/news/out.going files for all your sites did/didn't queue up those three messages as appropriate. IMPORTANT: Remember to do a "ctlinnd reload newsfeeds x" command every time you update your "newsfeeds" file! Finally, for checking out changes to newsfeeds, I've found "ctlinnd checkfile" handy. ------------------------------ Subject: Cookbook example setting up NOV ("overchan"). Christophe.Wolfhugel@grasp.insa-lyon.fr (Christophe Wolfhugel) (with many modifications from Tom Limoncelli) writes: Step 1: Upgrade to INN 1.4 or higher: Most of the bugs in 1.3 were related with overchan. In fact, the reason why many people used 1.3 without any problems was due to the fact that they were not using overchan (and didn't hit on some of the bugs that appeared for SVR4 users, all of which were fixed in 1.4) Step 2: Make sure INN is working. Get everything else working before you try to get overchan to work. You'll only confuse yourself. Step 3: Ponder if you have enough disk space. NOV uses up an additional 10%-20% of your news spool. This is a good 100 Mb if you have a full feed. The real space savings come when you delete your separate databases for trn, nn, and tin and use one unified database. All serious newsreaders will have NOV support soon. Step 4: Edit overview.fmt (it's in the $INN/site directory, or you can edit it where it was installed, in /usr/lib/news ) to include "Xref:full" as the last line. (i.e. uncomment out the last line). Step 5: Add this entry to your "newsfeeds" file. overchan gets it's data from a special feed. # This feeds header data to NOV: OVERVIEW!:*:Tc,WO:/usr/local/news/bin/overchan Read the "newsfeeds" man to make sure you understand what you've just done. Step 6: (optional) To create the original database: (run this as "news") % /usr/local/news/bin/expireover -a -s If you skip this step, access will be slow for articles that came in before you started "overchan". This is not a problem. You will get a lot of warnings in your "news.daily" output until you have received at least one new article in each newsgroup. [ Note: "a lot of warnings" means one for every newsgroup. This can make your news.daily report >6000 lines. The lines will all look like: overchan cant open clari/local/washington/.overview, No such file or directory overchan cant open clari/local/sfbay/.overview, No such file or directory overchan cant open uc/news/.overview, No such file or directory ] Step 7: Change the invocation of news.daily: In the crontab file for "news", edit the "news.daily" line to be something like: news.daily delayrm expireover (the expireover is required if you use overchan) Step 8: Inform your users that you now support "NOV, the News OverView database" and suggest that people switch to newsreaders that use newsreaders that are compliant with the Overview format. Step 9: You are done. Step 10: In a few weeks, drop support for mthreads, nnmaster, etc. (assuming you've upgraded to replacements that use Overview) ====================================================================== STARTUP PROBLEMS AND SYSLOG MESSAGES ====================================================================== ------------------------------ Subject: syslog message: ME internal no to group Nelson Minar <nelson@reed.edu> discovered the hard way that: If you set MERGE_TO_GROUPS to "DO", You have to have a "to" group listed in your "active" file or you will get the above syslog message and innd will not start. ------------------------------ Subject: syslog message: ME cant sendto CCreader bytes 4 No such file or directory (Rich Salz replies:) It usually means that some ctlinnd command timed out and gave up before innd could get around to replying. Always a problem with datagrams. :-) Usually not a problem in real life however. In INN1.3, the timeout stuff is handled better so most of these should go away. ------------------------------ Subject: syslog message: ME internal no control and/or junk group You must have a newsgroup named "control" and a newsgroup named "junk" for innd to start. ------------------------------ Subject: syslog message: ME bad_newsfeeds no feeding sites (Rich Salz replies:) The syslog message is telling you that you are not feeding news to any sites. You have to have at least one feed. (You may consider this to be a bug, it's just that I'm too lazy to make everything work right if you don't have any newsfeeds.) Until you go into production and start feeding sites, add a line like this: dummy-feed:!*:: ------------------------------ Subject: Why do all these "readclose" messages show up in my syslog? Chris Schmidt <cs@germany.eu.net> says: The "readclose" message indicates that a remote connection to your server was not correctly terminated with the server-command "quit". This can have two reasons. First the line your feed uses to connect to you might be instable so that the connection drops every now and then. Solution: either ignore theses messages or find out why the line is unstable. The second reason for these messages could be a missconfigured client-program at your feed. This means the program (e.g nntplink) does close the connection without sending the "quit" first. If you configure a lower number for the exit-timeout (-e) than the close-timeout (-C) in nntplink then exactly this will happen. Solution: ask your feed to fix its nntplink-setup. Let me repeat that: If you are using "nntplink" your -e value must be higher than your -C value. ------------------------------ Subject: "File exists writing symlinking article file -- throttling" QUESTION: I'm running INN 1.4, and the server throttles itself, saying "File exists writing symlinking article file -- throttling". Why? I have no clue, other than to note that the message is being emitted while innd/art.c tries to link a crossposted group. ANSWER: Innd wrote the article to comp/foo/123 and then tried to symlink it to alt/bar/128 and found that the symlink failed with errno == EEXIST. This generally only happens when your active file does not match your file/directory use. The two most common cases of that are: Trying to use MMAP on Ultrix Trying to use MMAP on Linux Some strange interaction with tind. If you are using Ultrix or Linux, turn off MMAP. You don't have a choice in this. The Ultrix mmap() function does something completely different than the Sun/BSD mmap() function. The Linux function gives you some of the functionality that Sun/BSD mmap() function has, but not enough. (The Linux people expect to have it fully up to spec eventually.) At least one person has reported problems with ICL DRS6000 SVR4 Unix when using MMAP. Try turning off MMAP if you find problems. It has been reported that tind writes to the active file and this confuses innd (innd assumes it is the only process writing to the active file). If you are using tin, upgrade to the newest version of tin which can read the overview (NOV) database instead of the "tin" database. To fix the active file (which may be corrupted), make sure nobody else is writing to the active file, then do ctlinnd renumber '' to get things synchronized again. ------------------------------ Subject: "cant fopen <newsgroup>/.thread No such file or directory" Q: nnrpd logs "cant fopen <newsgroup>/.thread No such file or directory" hundreds of times a day although I installed trn-3 and maintain an overview database. Why doesn't trn use overview files instead of mthreads data? A: trn-3 tries to open .thread files (or use XTHREAD) first because $spooldir/db.init still exists. Delete it. ====================================================================== MISC QUESTIONS ====================================================================== ------------------------------ Subject: Can I edit my configuration files where they are, or do I have to edit them in $INN/site ? Technically, you should edit those files in the $INN/site directory, but then typing "make" all the time becomes a grind. I found that I was constantly forgetting to type "make" and then I couldn't figure out why my changes weren't doing anything. The alternative is to edit things in place and let the install procedure complain. It will error out on the file, and you can copy that file to $INN/site then "make" again. ------------------------------ Subject: syslog: ME cant nonblock 15 Operation not supported. I get the following "syslog" message in /var/adm/messages: Dec 2 20:40:04 venus innd: ME cant nonblock 15 Operation not supported Answer: (from paulr@umbc4.umbc.edu (Paul Riddle)) It turns out that this is happening because /usr/spool/news on the machine running innd is an NFS-mounted filesystem, and innd is trying to do an FIONBIO on my feed file, which is under /usr/spool/news/out.going. (tal@warren.mentorg.com adds:) All news transports (INN, C news, B news) want the spool partition to be local. Newsreader can read from an NFS mounted partition without any problems but innd should only see local partitions. NFS has a blatant disregard for many of the file semantics that are needed for a good news implementation. If you don't agree, please feel free to prove the authors of B news, C news, and INN wrong. Include source code. :-) Systems without unix-domain sockets sometimes see this error. ------------------------------ Subject: Suddenly my active and history files are owned by root! rc.news runs from root. After that, everything else should run as news. More specifically, it sounds like you've run news.daily as root by mistake. Make sure all your cron jobs run as news and you'll be fine. If you have an old "cron" system, you might consider replacing yours with one of the many public domain replacements. If you can't create a different "crontab" for each user, the idiom is something like: 0 * * * * * su news -c '/do/this/as/news' ------------------------------ Subject: What can I do if I can't purchase the O'Reiley And Associates book on Managing Usenet? Hold a fundraiser? Seriously, this document will help you some. HOWEVER many people have thought that the install.ms doc was incomplete but then re-read the "First Time Installation" portion and were amazed how good it was. Personally, I've been a newsadmin for too long to be able to know if it would be good for beginners. ------------------------------ Subject: Getting an active file. > In appendix IV, the reader is told that "the easiest way to > [find out which groups to create] is usually to ask [your newsfeed > site] for a copy of their active file, and for you to add the entries > of the groups that you're interested in." It would have been nice to > get instructions on where this active file lives, and how to create > the new groups, without digging through manpages (I still haven't > found out what the proper path and incantations are. How are these > commands issued? As shell commands? As news articles?). If your neighbor doesn't know where his/her active file is, you should look for another neighbor. "man active" will tell you on the first line. Different sites put it in different places. The man page should tell you where it is. Here is how you zap someone else's active file to make it ready for re-use: sed <active.old >active.new \ -e 's/^\([^ ]*\) [0-9]* [0-9]* \([^ ]*\)$/\1 0000000000 000000001 \2/' ------------------------------ Subject: How do I use nntplink with INN? First of all, I don't personally recommend using this program. I feel that it is a gimick. However, if you decide to join the INN Instant Propagation Party (INN-IPP), I suggest that you first run the feed using traditional methods for a month so that you make sure you are used to INN and make sure that the feed is properly functioning. Once you're ready, here's a cookbook example of an newsfeeds entry using nntplink. netcomsv.netcom.com\ :*,!junk/!ParcPlace\ :Tc,Wmf,S1024:/usr/local/news/bin/nntplink -i stdin netcomsv.netcom.com Others have reported that Tc,Wnm works. Since I don't use this program I can't validate which is correct. INN 1.2 users should have an explicit S value (i.e. S1024 or S16384). Without it innd 1.2 can choke and lose data if the receiver is jammed. (fixed in INN 1.3). Ian Phillipps <ian@unipalm.co.uk> notes some criteria for using nntplink rather than nnptsend: > (1) If you have more than one backbone feed, you can save a lot of > bandwidth, without risk, if you use nntplink (less duplication of > articles over nearly-parallel paths). > (2) More important, if you have a large number of feeds, nntplink > permits them to be fed simultaneously with the same articles. No big > deal, until you think of the what's going on in the pagedaemon and the > disk cache. > A "ps uaxr" rarely catches nntplink in the act ("D"), despite my having > 17 of them last time I counted. Our biggest outgoing newsfeed delivered > 16398 articles yesterday, using a total of 380 seconds CPU on a Sun > IPC, and no disk time :-) ------------------------------ Subject: How do I use newsgate with INN? anselmo@nic.near.net (Ed Anselmo) reports: > Can someone who's running INN and the news2mail program in the > newsgate package send me a sample of the entry in the newsfeeds > file for a typical bidirectional gateway? Thanks! This assumes you configured "mail2news" to put a distinctive host-gateway name, like "news-mail-gateway" in your Path: headers. In /usr/lib/news/newsfeeds: ==================================== ## ## NEWS/MAIL GATEWAY ## tech-gateway/news-mail-gateway\ :ne.nearnet.tech\ :Tp:/usr/lib/newsbin/news2mail \ nearnet-tech nearnet-ops nearnet-tech-request nic.near.net %s ==================================== On the mailing list side, one of the recipients is: post-nearnet-tech: "|/usr/lib/newsbin/mail2news -n ne.nearnet.tech -o 'NEARnet News/Ma il Gateway' -d ne" ------------------------------ Subject: Other cron jobs. Once a night you should run the "news.daily" script which will expire old articles, run the daily reports, etc. It should run as "news" and look something like this: 40 23 * * * /usr/lib/news/bin/news.daily delayrm If you get news feeds via UUCP, you might want to add this cron job (also as "news") which checks to see if any batches arrived while innd was down and processes them. 20 * * * * /bin/rnews -U ------------------------------ Subject: After a crash. "What do I do after a system crash?" INN handles crashes pretty well. If there are any problems they get cleaned up by the nightly expire. About once a month you might want to run "makehistory -bu" to look for "lost" articles. Check the man page for "makehistory" for more information. ------------------------------ Subject: Debugging someone that is feeding you. David Myers <dem@meaddata.com> suggests that if a neighbor complains that their feed to you doesn't work: (1) make sure they've read the man pages, and (2) have them send a copy of their newsfeeds file. ------------------------------ Subject: Feeds suddenly can't connect anymore! Q: How come feeds tell me they can't connect to me any more? A: When innd starts up it reads the hosts.nntp file and looks up the IP addresses for all the entries mentioned there. The problem is that this data is dynamic (sometimes people change IP addresses), and innd never goes back to check. If your system stays up for days and one of your feeds changes their IP address, innd will reject them. Rich plans to handle this in INN1.5, but for now you might find it useful to do a "ctlinnd reload hosts.nntp" out of cron every day or so or when you notice there's a problem. ------------------------------ Subject: How does the "ME" line interact with the other lines? > I'm still a little confused about the ME line's second field. The man page as of INN 1.3 is much more clear on this. Basically, the second field of the "ME" line specifies the default for the rest of the feeds. Otherwise, it isn't used. The "active" file declares which newsgroups you accept and don't accept. Here are some examples: ME:!*::: foo:!junk:... --send no newsgroups ME:*::: foo:!junk:... --send all newsgroups except junk ME:!*::: foo:*,!junk:... --send all newsgroups except junk By the way, generally you do not want to send "junk" or "control" to your neighbors. ------------------------------ Subject: Why doesn't nntpget work? The nntpget in INN 1.2 doesn't work. Period. Upgrade to the latest version of INN. ------------------------------ Subject: More about the "to.*" groups (Thanks to jmalcolm@sura.net (Joseph Malcolm) for supplying these answers.) >1) Why did my local INN act on the sendsys posted to to.neighbor? to.* groups aren't magic to INN. Your system received the message, it acted on it. >2) Why did my neighbor send the cmsg to all of his neighbors? See 3. >3) Is is related to having the "control" group in our newsgroups patterns? Yes. > The INN docs say you probably don't want to do this, but they don't say > why. Actually, they do. This is from innd(8): Sites may explicitly have the ``control'' newsgroup in their subscription list, although it is usually best to exclude it. If a control message is posted to a group whose name ends with the four characters ``.ctl'' then the suffix is stripped off and what is left is used as the group name. For example, a cancel message posted to ``news.admin.ctl'' will be sent to all sites that subscribe to ``control'' or ``news.admin.'' There is also a pointer to this in newsfeeds(5). > But I still need it in my active file, right? Yes. ------------------------------ Subject: I'm getting groups sent to me that I don't want. Tell the system administrator(s) of the machine(s) that feed news to you to stop sending those groups. There is no other way to do it. (In B or C News, the groups would end up in junk; at least with INN they are not taking up space. You should compile with WANT_JUNK set to DONT). If the people that feed you use B news or C news, remember that they don't use a "newsfeeds" file. They use a file called "sys" which has a completely different format method for listing newsgroups. ------------------------------ Subject: When my feeder connects, I get articles but they don't take what's waiting for them. I hate to say this, but this really shows that you haven't RTFMed very much. News is not bidirectional (it's like SMTP, not UUCP). If you want to send things out you will have to make sure that you run send-nntp or nntpsend from cron. nntpsend is easier and elsewhere in this document there are cookbook examples of what to add every time you set up a new feed. ------------------------------ Subject: Directories are being created with wrong permissions. > Question: >When I received news for /var/spool/news/foo/bar for the first >time, the directories got created: > ># ls -lgR foo >total 1 >d-wx-w-rwx 2 news news 512 Feb 9 00:03 bar/ > >What did I do wrong? > >## Mode that directories are created under. >#### =()<GROUPDIR_MODE @<GROUPDIR_MODE>@>()= >GROUPDIR_MODE 2775 Answer: You forgot a zero in front of this number, for the C compiler to interpret it as octal instead of decimal. ------------------------------ Subject: Warnings to people that must set HAVE_UNIX_DOMAIN to DONT Disclaimer: First of all, if you have to set HAVE_UNIX_DOMAIN to DONT, YOU HAVE TO SET IT to DONT. It's not a choice you can make, it's a description of the operating system that you purchased. If you wrongly set this variable to DO your system isn't going to work *at* *all*. When you use the POST via NNTP, you are talking to nnrpd. nnrpd cleans up your headers, adds the missing headers that it is allowed to add, checks whatever it checks, and then submits the finalized version to innd. How does it talk to innd? If you have HAVE_UNIX_DOMAIN set to DO, nnrpd opens a Unix domain socket and sends the text. At this point it is talking to innd somewhat like ctlinnd does. innd can trust that the post isn't forged since it is coming from a program trustworthy enough to get to the socket (which isn't much). If you have HAVE_UNIX_DOMAIN set to DONT, it has no choice but to open a socket to port 119 and issue the "IHAVE" command, and send the text that way (just like a remote newsreader). This means that innd (not another nnrpd) has to be at the other end of the pipe. If it opens the connection and sees a "nnrpd" you're hosed and you get "441 480 Transfer permission denied". To get innd to not hand off the connection to a nnrpd process, you must have the host's name in the hosts.nntp file. (don't forget to do "ctlinnd reload hosts.nntp") If you have your host's name in the hosts.nntp file, then any newsreader running on your nntphost must be "INN-aware" (i.e. that they issue the "mode reader" command) or they must read news via the file system instead of NNTP. There is no way around this right now. - * The End * - 
