11.1 Are Highwinds products Year 2000 compliant?
(24/11/98)
1 Introduction
1.1 Introduction (16/10/98 Highwinds Support
) This
FAQ assumes that the reader is familiar with the basic operation of Highwinds products.
It is not intended to replace the documentation, but instead, is intended to be
used as a supplement and provide answers to common and advanced questions. This
FAQ also assumes that you are running the latest version of Highwinds products.
Features and hints documented in this FAQ may not work for older versions. Any
comments or questions for the FAQ are welcome. Please direct correspondence to Highwinds Support.[top]
1.2 Getting Support (16/10/98 Highwinds Support
) All
of Highwinds's products are FULLY supported. We are happy to provide pre-sales support
and have training and support options ranging from EMAIL support to 7x24 Network
Operations Center available for purchase.
For software support or questions about software mail Highwinds Support.
For Usenet access support questions mail service@highwinds-software.com.
[top]
2 Hardware
2.1 General Recommendations
(17/09/98 Highwinds Support
) An important
consideration to take into account when purchasing hardware is that most people
purchase a much faster CPU and, for Cyclone, much more disk space than they need.
It is often better to spend more money on memory. For Cyclone, scalability may
not even be an issue. If you are using Cyclone with less than 20 feeds, almost
any hardware will be adequate. For Typhoon and Breeze, since you have to archive
and manage articles, your machine will require more resources.[top]
2.2 CPU (17/11/98 Highwinds Support
) CPU
power is not usually of critical importance. Memory, disk, and network usually
become bottlenecks before the system becomes CPU bound. Using filtering, subscription
directives, active files and other features will all increase CPU usage. With
Typhoon and Breeze, if you turn on compression, you can radically increase the
demand for CPU power. Our recommendation is that you start out with a modern CPU
and only consider adding CPU if you are running a very large system or notice
obvious CPU bottlenecks.[top]
2.3 Memory (16/10/98 Highwinds Support
) There
are no hard rules for how much memory a Highwinds based server requires. As always,
more memory is better. It is also a good idea to have swap space spread across
multiple drives. As a general rule, 256k to 1MB for each connection is more than
sufficient. Exceeding this rule is fine, as long as it is within reasonable limits.
Using certain history options such as "-cache" will also increase memory usage.
In Typhoon and Breeze, high compression levels can also consume extra memory.
Remember, Highwinds products will use all available memory for caching and buffering
multiple parallel disk and network I/O's. You don't necessarily NEED a ton of
memory, but, having extra makes things perform better.[top]
2.4 Disk (16/10/98 Highwinds Support
) More
spindles and more controllers are always helpful. The disk configuration regulates
not only the speed in receiving and transmitting articles, but also provides an
upper limit to how much data you can store. SCSI is highly recommended, and systems
with Fiber Channel, UltraSCSI or Wide interfaces will be able to better utilize
their disks. For Typhoon and Breeze, using RAID 0+1 or RAID 5 subsystems is recommended.
JBOD across many drives has also proven very successful on large systems. In all
disk scenarios, due to the nature of USENET, you should optimize your disk subsystem
for many small random I/O's.[top]
3 O/S Configuration
3.1 Disk Configuration
(16/10/98 Highwinds Support
) For
most sites, even simple disk layouts should work fine. However, if you are running
a large or busy site, you should review the notes in the Cyclone Guide
or the Reader Configuration and Operation Guides. As mentioned earlier, striping and isolating subsystems
in both Cyclone and Typhoon are good ideas. In Cyclone, separating the history,
logs, and outgoing queues from the spools objects is encouraged. For Typhoon,
spreading and isolating the overview databases, spool objects, and overview cache
objects will result in increased performance.[top]
3.2 Swap Space (17/11/98 Highwinds Support
) Spread
your swap space across multiple drives. Our advice is to make 1 - 1.5 times your
physical memory available for swap space. Reference the "swap" and "mkfile" UNIX
commands if you need help adjusting swap location on your system.[top]
3.3 Solaris specific tuning
(02/12/98 Highwinds Support
) The
/etc/system file allows you to modify kernel values. The system must be rebooted
for any changes to take effect. Some suggested tuning parameters for Cyclone include:
* How often to check buffers before flushing to disk
set autoup=360
* time (in seconds) after which buffers will be saved to disk
set tune_t_fsflushr=30
The previous two settings will cause the OS to flush buffers to disk less often.
Priority Paging is deprecated in Solaris 9 and above.
By default in Solaris 8 and below, prioirty paging
is not enabled. To turn it on add the following lines to your /etc/system and
reboot.
* Enable Priority Paging
set priority_paging=1
In addition, if you have modern disk hardware to hold your swap, you should probably
increase the maxpgio kernel variable to increase the performance
of the Solaris page stealing daemon.
set maxpgio=100
or
set maxpgio=200
[top]
4 Software Configuration
4.1 Can I share spools between two machines with
NFS? (09/11/98
Highwinds Support )
No. Sharing spools via NFS was abandoned for two reasons. First,
it doesn't work with large files like our spool objects. When one machine updates
an open file, a huge amount of time may pass before other machines that have that
file open see the changes. Second, the desire for spool-sharing came from the
behavior of other News systems, which would max out a processor or memory while
the drives were still idle. Typhoon has the opposite dynamics: generally the drives
are the first thing to become a bottleneck, so the shared-spools scenario loses
its load-sharing benefits.Our chaining mechanism does allow a type of spool sharing,
where one box advertises the contents of the other's spools. This can be used
for load sharing, by setting up "front-end" boxes which only contain the most
popular or most recent groups, and a big "back-end" box which has a larger spool.
This is described in the Configuration Guide.
[top]
4.2 How should you configure the software to work
with NFS mounted drives? (13/04/99
Highwinds Support )
All of Highwinds's products have the "-nfs" command line option.
This option causes the software to optimize its disk accesses for NFS mounted
data. In addition, we strongly suggest that you use multiple mount points for
system components.Because of limitations in various UNIX systems, multithreaded
I/O requests can get bottlenecked through a single NFS mount. For that reason,
we recommend that you mount system components (spool objects, overview databases,
history, etc.) via individual mount points REGARDLESS of whether or not they are
from the same NFS server. By doing this, you give the kernel multiple parallel
access channels and eliminate bottlenecks. In addition, if you are running under
Solaris and are producing high NFS loads, you may wish to adjust the nfs_client_threads
kernel parameter. [top]
4.3 Can I change the size of my spool objects without
losing data? (23/10/98
Highwinds Support )
You can freely increase the size of your spool objects by increasing
their "Kilobytes" value in typhoon.conf of breeze.conf and restarting the server.
The server will take care of further extending the file and no messages will be
lost.You can also decrease the size of your spool objects by decreasing their
"Kilobytes" value, truncating their spool files carefully, and restarting. Of
course articles will be lost, and there is no guarantee that the lost articles
will be the oldest ones because spool objects do their own internal allocation.Never
change a spool object's SpoolNumber. Doing so causes the loss of every article
in that spool object. [top]
4.4 How can I create or add spools quickly?
(18/11/98
Highwinds Support )
There are several ways to create spools when installing a server
or adding spools to an existing server. If you are installing a server and have
many spools to create you should start by putting only one spool object in your
conf file. Run bin/start which will create the first spool file and start the
server. Once the server is running, modify your conf file and add the other spool
objects to it, then run "bin/validate -spools". This will create the other spool
files while the server fills up the first spool. Once the other spool files are
created run "bin/restart" to start utilizing all spools. If you are adding spools
to an existing server you can simply modify your conf file, and run "bin/validate
-spools" with the server running. Once the new spool files have been created run
"bin/restart" to start utilizing all spools. Since spool files are regular files,
another option to create them is to copy existing spool files to the locations
specified in your conf file. This should save some time on systems striping a
spool partition between multiple disks. You can NOT copy spool files between machines
of different architectures, including NFS mounted spools.[top]
4.5 How can I delete spools?
(16/11/98 Highwinds Support
) To delete
spool objects, you can remove the spool objects in question from your conf file
with the server running, and then restart the server. After the server has been
restarted, remove the spool files by using "rm".[top]
4.6 What size history should I use?
(04/02/99
Highwinds )
The number of article ids you wish to keep in your history will
determine what size you should use.
Cyclone history database sizes
#of article ids amount of disk space
4 million 67 MB
8 million 112 MB
12 million 168 MB
16 million 218 MB
24 million 319 MB
By default Typhoon and Breeze create a history to store 8 million article ids
using 218 MB. Be default Cyclone creates a slightly different format history to
store 8 million articles using 112MB. Memory requirements are approx. 8-16MB for
any size history database. History databases are self-expiring and are a fixed
size. This means you do not need clean or purge the history, and the history database
will never require more space than when it is created. You will have to restart
your history if you change the size of your history database. Using the -cache
option will store large portions of the history database into memory. This will
increase performance on systems with lots of free memory.[top]
4.7 How big should my overview caches be?
(09/03/99
Highwinds Support )
First, the key to overview caches is specificity, not size.
Even if you only have one disk on which to place your overview caches, four four
small caches with specific Subscription and FilterSubscriptions (say alt.*,
alt.binaries.*,
rec.*, and everything else) will be much faster than one larger overview
cache with a subscription of *.As a matter of fact, an overview cache with
"Subscription *" does not help performance at all.Now, as for size: as a rule
of thumb, you need to keep 1/20th (5%) as much overview cache space as you have
spool space for a given hierarchy. For instance, if you have 10Gb of rec.*, your
overview cache for rec.* should be 500Mb or so.If you want to make a more careful
calculation: overview caches are simply filled with XOVER reply lines. Let's say
you have 10Gb of rec.* and know that you have 2.5 million articles in that 10Gb
from looking up the oldest articles in a popular group in your logfile and seeing
what's arrived since. Now, assume you estimate that most rec.* articles have XOVER
reply lines that are about 250 characters on average. Then you need 2.5M * 250
= 625Mb of overview cache for rec.*. Of course, you are always free to make overview
caches smaller or bigger if you desire: smaller caches will mean that only XOVERs
of newer articles are cached, while larger caches will mean that the cache has
extra room. [top]
4.8 What are Path lines and what can I do with them?
(18/11/98
Highwinds Support )
Path lines are found in the header of news articles usually showing
what servers an article has been through and can be used to control news routing.
Usenet sites can use path stamps to determine what articles to offer other servers
as well as reject from them. Highwinds news servers provide several simple but powerful
options to handle how the server will stamp articles. -masquerade option: Normally
the server stamps incoming articles with the full hostname of the server. Using
the -masquerade option will allow you to stamp an alternative path line on articles.
-alias option: The -alias option lets you specify multiple aliases which the server
will check against the Path header of incoming articles. The server will then
add the missing stamps to the path line for each alias not found. For example
a server started with the options: "-masquerade newsfeed.site.com -alias
site.com!news.site.com"
IN : border.site.com!hello!usenet
OUT: newsfeed.site.com!site.com!news.site.com!border.site.com!hello!usenet IN
: border.site.com!site.com!hello!usenet
OUT: newsfeed.site.com!border.site.com!site.com!hello!usenet IN : border.site.com!site.com!hello!news.site.com!usenet
OUT: newsfeed.site.com!border.site.com!site.com!hello!news.site.com!usenet[top]
4.9 How can I run multiple servers on the same machine?
(16/11/98
Highwinds Support )
By using the "-interface" command line option you can run your server
on a specific network address as opposed to listening to all addresses. This allows
you to run more than one server on the same machine, on the same or different
ports. As always, you can NOT share spool file between different servers, even
if they are on the same machine.[top]
4.10 How do I setup a virtual news server?
(26/02/99
Highwinds Support )
Typhoon servers can be setup to run as a collection of multiple
"virtual" news servers for different customers. The virtual servers are configured
by defining feed objects in your feeds.conf. All the standard feed object directives
apply for Virtual servers. It is simplest to set up a virtual server when you
can specify what view of the news server a user should see based on that user's
hostname or IP address. In that case, you simply create different feed objects
which specify the IncomingHostNames of each class of user, the Subscription/FilterSubscription
of the groups that these users can see, and optionally a WelcomeMessage that the
server issues at the beginning of each connection.Here is an example of simple
virtual servering:
# Employees can read anything.
<Feed>
IncomingHostNames *.mycompany.com
WelcomeMessage Welcome to My Company's News Server.
Subscription *
FilterSubscription !*
AllowReading Yes
AllowPosting Yes
</Feed>
# Students can read non-alt groups but can't post.
<Feed>
IncomingHostNames *.k12.mystate
WelcomeMessage Welcome to the K-12 news server.
AllowReading Yes
AllowPosting No
Subscription *, !alt.*
FilterSubscription alt.*
</Feed>
You can always use DNS CNAME's to alias different names to the same actual server.
In addition, in Typhoon version 1.1.9 and later you can tell different classes
of users to use different names for the news server, and map these names to network
interfaces on the machine. These interfaces can be either physical or virtual.
Then, you use multiple feed objects with different IncomingInterface directives
to specify a Subscription/FilterSubscription for each class of users.If you need
a virtual servering setup more complicated than any of the above, you can do almost
anything with program-based authentication, including the host- and interface-based
distinctions above and even per-user permissions that vary based on what time
of day the user connects. See the documentation
on program-based authentication for details. [top]
4.11Why do I need Subscription and FilterSubscription?
(06/29/99
Highwinds Support )
FilterSubscription wouldn't be necessary if every article were
posted only to one group.
But with the possibility that an article might be crossposted both to
comp.lang.c++
and to alt.binaries.evil, you need FilterSubscription: let's say you don't want
to let your programmers see anything outside the comp groups, and they try to
get sneaky and get people to crosspost between the comp group and alt group mentioned
above.
With just "Subscription comp.*" or even "Subscription comp.*,!alt.*" on the programmers'
feed, then an article with "Newsgroups: comp.lang.c++,alt.binaries.evil" is _still_
visible to the programmers because "comp.lang.c++" matches the subscription.
However, with "Subscription comp.*" plus "FilterSubscription alt.*", the article
is invisible to the programmers' feed.
An article matches if:
1) SOMETHING on its Newsgroups line matches the Subscription line, and 2) NOTHING
on its Newsgroups line matches the FilterSubscription line.
Some people like to say that FilterSubscription creates "poison" groups.
[top]
4.12Does Cyclone ever delete/clean-up/destroy spool objects?
(07/07/99 Highwinds
Support )
No. When Cyclone first starts up, it creates, allocates, and consumes
the entire specified amount of disk space indicated in the cyclone.conf file.
The collection of large files that are created are called "spool objects" and
they are never deleted by Cyclone. As stated in the documentation, we recommend
that you spread your spool objects among the disk drives on your system. That
being said, Cyclone does do periodic maintenance and deletion of old backlog files
in sub-directories of the "spool" directory off of your Cyclone installation area.
For that reason, you should make sure that your spool objects are not in sub-directories
of the "spool" directory. [top]
5 Feed Configuration
5.1 Peering (16/10/98 Highwinds Support
) To ensure
a high-quality incoming news stream, we recommend that you "peer" (exchange news)
with a number of high-quality news sites. For most systems, 3-4 good peers will
guarantee an excellent feed. If you are running Cyclone, peering with other Cyclone
sites will result in better bandwidth usage for all parties involved. Cyclone's
adaptive streaming engines "lock on" to each other and result in faster and more
efficient article transit.[top]
6 Adding/removing groups, articles and the active file
6.1 Where can I get a complete up to date active
file? (16/11/98
Highwinds Support )
ISC is the official maintainer for the "Big-8" set of newsgroups
and keeps up to date versions of both active and newsgroups files. They can be
found at: ftp://ftp.isc.org/pub/usenet/CONFIG/[top]
6.2 How should I add/remove groups from the active
file? (16/11/98
Highwinds Support )
There are several ways to add, remove and modify groups in the active
file. Do NOT edit the active file while the server is running because the server
frequently updates it and it can become corrupted. To add, remove or modify groups
in the active file while the server is running you should use the active.control
file. Refer to the server documentation on using the active.control file. When
the server if off you can safely edit the active file by hand or with a script.
Be sure to keep the high article numbers or you will not be able to access previous
articles. You may be able to make a temporary copy of the active file, quickly
modify the copy, and move it the to current active file with the server running
though this is not recommended.[top]
6.3 How do I delete specific articles?
(10/02/99
Highwinds )
Create a file called "cancel.control" in the same directory as your
active file. Put either a message-ID like "<foo@bar.baz>" or something like
"alt.test:2233" on each line. Then, run bin/reload. The server will cancel the
listed messages and delete the cancel.control file. Make sure this file is readable
and writable by the news user.[top]
6.4 How do I stop propagating unapproved articles
without using a large active file? (19/11/98 Highwinds Support
) In Cyclone
there is a clever way you can stop propagating unapproved articles in moderated
newsgroups without using a large active file. Make a small active file which contains
only moderated newsgroups in it. Then specifying "DefaultAllowJunk True" in your
cyclone.conf will results in the propagation of all articles not in the active
file.[top]
6.5 What are the "Big 8" newsgroups?
(16/11/98
Highwinds Support )
The "Big 8" are the original main eight news hierarchies considered
to be comp.*, humanities.*, misc.*, news.*, rec.*, sci.*, soc.*, talk.* Despite
large amounts of groups and articles alt.* is not considered to be one of the
"Big 8" [top]
6.6 I added a newsgroup to Typhoon/Breeze and it
appeared, but it's not getting any articles! (17/09/98 Highwinds Support
) Check
the XrefAction setting in typhoon.conf or breeze.conf. If it's "XrefAction Parse",
then you probably just need to add the group to the server that's numbering articles,
which may be a Cyclone or another Typhoon. Find the box that has "XrefAction Generate"
in its cyclone.conf or typhoon.conf, and add the group to that machine's active
file.[top]
6.7 How can I create a local group?
(19/11/98
Highwinds Support )
Setting up local-only newsgroups is relatively easy in Typhoon and
Breeze. For instance, to create an "internal.*" newsgroup hierarchy which can
be posted to be which won't exchange those messages with the outside world, you
need to:
- add the new groups to the active file, either with the server shut down or
via the active.control/reload mechanism.
- add internal.* to the UpstreamFilterSubscription in typhoon.conf. This prevents
internal.* articles from going out.
- add internal.* to the DefaultIncomingGroupFilter in typhoon.conf. This prevents
unhappy former customers at other ISPs from flooding your internal discussion
groups from their new home.
- (Typhoon only)add internal.* to the FilterSubscription for any outgoing
feed objects you have in feeds.conf.
[top]
7 Statistics, Logfiles, and Monitoring
7.1 How do I produce statistics reports?
(09/11/98
Highwinds Support )
All Highwinds Software products produce detailed raw statistics files
that can be the basis of excellent statistical reports. To get started, take a
look at the scripts in the "tools" directory and the "README.stats" file in the
"etc" directory. In addition, a number of excellent graphical packages are available
on the Highwinds Software Contributed Software Web Site[top]
7.2 How can I measure the retention in my spools?
(04/02/99
Highwinds )
If you run with the Typhoon and Breeze with the -log option, or
Cyclone with the -loggroups options, you can use the Perl script found in
tools/log-report.pl
to process the logfile and tell you the retention in each of your spools. If you're
feeding multiple logfiles into log-report.pl, make sure to feed them in chronological
order for the most accurate results. Also, see the next question, "Why does log-report.pl
say that my spools are only 43% full?"[top]
7.3 Why does log-report.pl say that my spools are
only 43% full? (17/09/98
Highwinds Support )
Log-report.pl can only calculate spool capacity based on the logfile
it was provided. If your server's been up a long time but you rotate your logfiles
regularly, then the capacity reported will only be the amount of your spool consumed
since your last logfile rotation.[top]
7.4 What's the best way to rotate my logfiles and
analyze them? (17/09/98
Highwinds Support )
The logfiles generated by -log, -stats, and -paths are all flushed
and re-opened each statistics interval, or whenever bin/statsnow is run. The best
way to rotate your logfiles is:
- mv logfile logfile.old
- run bin/statsnow
- sleep 15 (to wait for the file to flush)
- do any post-processing, like compression or reports generation
You don't need to "touch" the new logfile.[top]
7.5 How do I setup my server to log to syslog?
(18/11/98
Highwinds Support )
It is usually sufficient to log "news.info" and "news.crit" priorities
to logfiles. Logging "news.info" will only be utilized if the server has been
started with the "-detailed" option. An example /etc/syslog.conf entry would be:
news.info /path/to/news/log/news.info
news.crit /path/to/news/log/news.crit
Fields are TAB separated but please check your syslog.conf documentation for exact
syntax. For sites running many servers, it may be beneficial to log all or just
high priority messages to a remote centralized host for monitoring. [top]
8 Maintenance
8.1 Hourly/Daily Cronjobs
(18/11/98 Highwinds Support
) An alternative
to using the "-update" value for specifying statistics periods, sites might wish
to handle stats through a cron script. A benefit of this is it allows you to restart
the server if the configuration has changed while still generating stats on the
hour. To do this, make a root cronjob for the top of the hour. The script should
check the modified time of the configuration files, and if they have been modified
in the last hour, execute the "bin/restart" script. Otherwise, if they have not
been modified, execute the "bin/statsnow" script. You might also wish to generate
reports from your statistics files after the stats have been written. Remember
to set the "-update" time to a large value (perhaps 2 hours in case the cron job
fails). Also, for security, make the scripts owned by root and read-only.[top]
9 Server and Client Error Messages
9.1 Why are my users not seeing articles or newsgroups?
(24/03/99
Highwinds )
Articles and newsgroups can appear to dissappear in a news client
as aresult of either news client issues or Typhoon configuration issues. To isolate
where the problem is occurring, first telnet into the news server to check for
the existence of the newsgroups/articles in question. This will help determine
if the problem exists in the news client or the news server.
Example:
telnet news.your.domain 119
group some.group.to.check
article [message-id|number]
quit
If the group exists, and there is traffic in the group, then the problem is likely
news client related. The solution for news client problems is to reset the last
read article counter. Article counters in news clients somtimes become corrupted
or de-synchronized from the article numbering sequence on the given news server,
causing articles and groups to "disappear". Ressetting the article counter will
resynchronize the client's article counter with the news server article numbering,
causing groups and articles to become visible. If the group does not show up on
the server, or contains no messages, then the problem is likely a configuration
problem in Typhoon. Check the Subscription and FilterSubscription directives in
the <Feed> objects in typhoon.conf. Also, edit bin/start to start Typhoon
with the "-log ../log" option and look in log to see which spool the articles
are being stored in. The format of the log file is documented in etc/README.stats.[top]
9.2 "signal fault in critical section"
(24/11/98
Highwinds Support )
signal number: 11, signal code: 1, fault address: 0x0,
pc: 0x0, sp: 0x6f383c68
libthread panic: fault in libthread critical section (PID: 14928 LWP 3)
stacktrace:
6fee3456
...
This error message is seen under Solaris when the thread library is old and in
need of patching. While in many cases the server will continue to run as if nothing
happened, this clearly indicates a major problem with the threads library and
the OS should be brought up to the current patch level as soon as is convenient.[top]
9.3 "Unexpected Response to IHAVE"
(22/02/99
Highwinds )
If your cyclone server is logging the following error:
cycloned[9935]: [10.0.0.10] Unexpected Response to IHAVE
<7665e064.3424442@some.com> Received: 500 Syntax Error or Unknown
Command. Dropping and Disconnecting.
Then the news server you are trying to feed articles to is not allowing your cyclone
server to feed it via IHAVE. If the news server is yours and is either typhoon
or breeze, add "AllowFeeding True" to its feed object for the cyclone server,
and run bin/restart. If the news server is run by another admin, you will need
to tell him/her to change their configuration on the incoming stream from your
cyclone server. They will need to know how to change this but it will likely be
an allow feeding or allow streaming option. A manual check you can run is to log
onto your cyclone server and telnet to the news port on the news server in question.
Type:
ihave <foo1234567@foo.com>
If you get a "335" response, the server will accept ihave feeds. If you get a
"5xx" response, the server is not configured to accept ihave feeds. [top]
9.4 Why do my users get a "Not in newsgroup" error?
(30/11/98
Highwinds Support
)
A form of the following error can be seen on the client software
of some modern news readers.
News Error!
News host responded: Not in a newsgroup
Perhaps the article has expired
<364B9DD7.393E488A@dtc.net> (229976)
Some modern news readers access newsgroups via multiple connections and will attempt
to open many connections to the News server. A downside to this is when a user
attempts to access a second article before the previous command has completed.
This can cause the news reader to retain header data and article number of the
current article when spawning an additional connection. Also if a user is allowed
one primary and one backup connection, and the news reader attempts to open another
connection the user may receive this error. This occurs most often when the user
is browsing articles within a single group.[top]
9.5 "430 No such article" with Message-ID but not
article number (07/12/98
Highwinds Support
)
This is just the result of history expiration. By default, Typhoon
runs with an 8-million article history. This is somewhere between half a week
and a week depending on the size of your newsfeed and whether you receive cancels.
In addition to its use during feeding, the history is used when looking up articles
by Message-ID. (The overview database is used for looking up articles by number.)
This means that articles more than a few days old generally aren't retrievable
by Message-ID with the default history size, but may be retrievable by article
number. If you want to keep a bigger history, you can add "-queue" to Typhoon's
command line in bin/start: for instance, you can get a 9-million article history
with "-queue 9".[top]
9.6 Why is my Typhoon logging "Article was unwanted
during POST. Received 435"? (24/03/99 Highwinds
Support
<Highwinds Suppor>) Typhoon
has two mechanisms that feed articles to other servers. The upstream mechanism,
controlled by the UpstreamHostNames directive in typhoon.conf, is used in feeding
locally posted articles out to the rest of the world. The outgoing feeding mechanism,
controlled by the OutgoingHostName directive if feeds.conf, is used in re-feeding
articles received by peers and also in sending out local posts.The upstream mechanism
assumes that it is the only method by which a post can make it to the outside
world. This message is the upstream mechanism warning you that your upstream host
claimed to have already seen a local post which you offered it. This probably
means that you have your upstream box set as both UpstreamHostNames and
OutgoingHostName,
and that the outgoing feeding got ahead of the upstream mechanism.Use the upstream
mechanism if you have a separate peering/transit box. On the Typhoon, set the
UpstreamHostNames to the name of your transit box, and don't set up an outgoing
feed to the transit box.Use the outgoing feeding mechanism if your Typhoon does
its own peering. In this case, you don't need to set UpstreamHostName.
[top]
10 External Programs and Filters
10.1 Why does my SPAM or POST filter get out of
sync and/or hang? (06/10/98
Highwinds Support <Highwinds Support>)
Since SPAM and POST filters operate on data from the NNTP stream,
they must be able to handle any type of input. A typical problem in these filters
is that the author simply checks for a ".\r\n" on a line by itself as the termination
of an article. This is not sufficient. The program author must also
ensure that the previous line was terminated by a "\r\n". More specifically, the
termination of an article is "\r\n.\r\n". This problem frequently occurs when
the author is using Perl for the program. To avoid this problem, we suggest that
you either keep track of the previous line or simply set the special line termination
variable like this:
$/ = "\r\n";
This code changes the Perl line-delimiter to "\r\n". By doing this, every time
a "line" is provided to the Perl program, it will be terminated by a "\r\n" and
not just "\n" (the Perl default).[top]
10.2 Why can't I post to Typhoon with GNUS?
(18/11/98
Highwinds Support )
GNUS uses incorrect line-termination when posting and terminating
articles: "\n" instead of "\r\n" as required in NNTP wire format. The following
patch to a GNUS routine fixes the problem. Thanks to Paul Southworth. Make these
changes to nntp-encode-text:
(defun nntp-encode-text ()
"Encode text in current buffer for NNTP transmission.
1. Insert `.' at beginning of line.
2. Insert `.' at end of buffer (end of text mark).
! 3. Terminate each line with CRLF [db3l]."
(save-excursion
;; Insert newline at end of buffer.
(goto-char (point-max))
(or (bolp) (insert "\n"))
;; Replace `.' at beginning of line with `..'.
(goto-char (point-min))
;; (replace-regexp "^\\." "..")
(while (search-forward "\n." nil t)
(insert "."))
! ;; [db3l] Replace LF at end of lines with CRLF
! (goto-char (point-min))
! (while (search-forward "\n" nil t)
! (replace-match "\r\n" nil t))
;; Insert `.' at end of buffer (end of text mark).
(goto-char (point-max))
(insert ".\r\n")))
[top]
10.3 Where do articles posted by my users get filtered?
(03/12/98
Highwinds Support )
This question usually comes up when a company's Typhoon server is
setup behind their primary Cyclone server. The following shows the cycle an article
takes when posted to the Typhoon server regardless if filters are being used:
- User posts article to Typhoon in "XRefAction Parse" mode,
- Article is run through Typhoon's post filter if present,
- Article is sent to Cyclone om "XRefAction Generate" mode,
- Cyclone accepts the article
- Article is run through Cyclone's spam filter if present
- Cyclone numbers the article
- Article is queued into outgoing feeds, including to Typhoon
- Typhoon re-accepts the article
- Article is put through Typhoon's spam filter if present
- Article appears to users
[top]
11 Statements and Disclaimers
11.1 Are Highwinds products Year 2000 compliant?
(24/11/98
Highwinds Support )
Yes, all versions of all products Highwinds Software has ever produced
are fully Year 2000 compliant. Highwinds products use 4 digit years and will properly
process and use dates before, during, and after the Year 2000 transition provided
the underlying operating system of the host machine is Year 2000 compliant. If
you would like a FAX'd copy of our official Year 2000 compliance statement, contact
Highwinds Support[top]
"Cyclone NewsRouter", "Typhoon NewsServer" and "Breeze NewsServer" are trademarks
of Highwinds Software
Copyright © 2006 Highwinds Software LLC All Rights Reserved