DNS
Waif
New to Unix, so pardon any idiocy. I have been playing with a SCO 5.0.5 OE
server for a couple months now, as time permits (which ain't often).
I am presently trying to clear up the following issue: The first time I try
to run a program off the SCO server via Windows95, the system takes 5-8
minutes to pull up the application, but subsequent attempts to run these
applications results in immediate access. I've tried this with VisionFS
Profile Manager (3.0) and with an ODBC link to our new manufacturing
software. Note that telnet and access to VisionFS shares through Network
Neighborhood does not appear to have this problem - both connect
immediately.
I posted previously and was told this was more than likely an issue with DNS
being improperly set up, which is highly believable.
So now I'm looking for resources on how to set it up, or more importantly,
advice on which type to set up. We have 1 SCO server (128MB Ram, Pentium
II 266, 6GB disk space). We are a small company with 25 desktop systems and
a single server. I can't justify spending the cash for another SCO box just
to act as a DNS server, so I want to run DNS off the one box we have now.
But I don't want to slow down this box (file server) too much either. I
assume I want to set up this box as a primary DNS server, but am concerned
that it will slow things down too much. Do I want a cache-only server
instead? Or maybe I don't need DNS at all??
Also, is there a way to check if components are already set up so I can
strip them and start over?
--
James Horvath
(jhorvath @ frabill.com)
Jim Richardson
James
DNS will become important as you connect to the outside world. If the
server and your pcs are only an internal network you can handle things
with host files and LMHOSTS.
Case 1: all internal you'll need a copy of hosts on each pc and on the
server. List all the units and their ip addresses.
Case 2: connection to the outside world. In network setup on each pc
enable DNS lookup and enter the address of the DNS server you currently
use.
But I'll bet your current problem is not just DNS. Can you give us a
little more information about what you're trying to do?
--
Jim Richardson
I like NT because it constantly reminds me of my daughter.
"Honest Daddy, I wasn't doing anything and it just broke."
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Tony Lawrence
> So now I'm looking for resources on how to set it up, or more importantly,
> advice on which type to set up. We have 1 SCO server (128MB Ram, Pentium
> II 266, 6GB disk space). We are a small company with 25 desktop systems and
> a single server. I can't justify spending the cash for another SCO box just
> to act as a DNS server, so I want to run DNS off the one box we have now.
> But I don't want to slow down this box (file server) too much either. I
> assume I want to set up this box as a primary DNS server, but am concerned
> that it will slow things down too much. Do I want a cache-only server
> instead? Or maybe I don't need DNS at all??
Maybe.
You can add all the Windows ip addresses to /etc/hosts, and
on the windows side create a C:\windows\lmhosts file
containing
xxx.xxx.xxx.xxx scobox #PRE
where xxx is the address of your sco box and "scobox" is its
name.
However, if you DO need DNS, and don't want to set it up on
your server (it really wouldn't be any intense load,
though), a couple of suggestions:
Setting it up on NT is easy if you are unfortunate enough to
have that in your mix.
Or, an old 386 running Linux would also be simple to set up,
and a lot more useful than NT.
But try the hosts/lmhosts route first..
>
> Also, is there a way to check if components are already set up so I can
> strip them and start over?
If you have a DNS server of any kind set up, you'll be
running named and there will be a file /etc/named.boot that
points named at the databases. On 5.0.5, it's been changed
to /etc/named.conf for reasons that must have been terribly
compelling :-)
If you don't have any of those, than you either have no DNS
or have a resolver only set up, which would be indicated by
having an /etc/resolv.conf (yep, no "e" on resolv) which
tells the machine where to get its DNS info from. See "man
named" and "man resolv.conf" for more info; I'd suggest
O'Reilly's "DNS and BIND" as a reference if you are just
getting your feet wet here:
http://www.amazon.com/exec/obidos/ASIN/1565925122/aplawrencescouni/002-4383974-7052064
--
Tony Lawrence (to...@aplawrence.com)
SCO articles, help, book reviews, tests,
job listings and more : http://www.aplawrence.com
Waif
You already have the basic machine specs - Intel Pentium-II 266, 128MB Ram,
Intel Pro+ 10/100BaseT NIC, 6GB hard drive, 8-port digi-board with nothing
attached yet. The network is a 10BaseT Ethernet TCP/IP LAN with a single
SCO Enterprise 5.0.5 (patched to the gills) file server with 25 Windows95
workstations. No remote access at this time. No workstations outside the
main office. The whole thing runs on Cat5 through 3COM Office Connect hubs.
SCO was installed by an outside vendor from whom we purchased a new
manufacturing software package. They know enough to get things running, but
aren't wizards by any means.
From the looks of it, I have a ton of stuff installed (most of which I'm
quite sure I don't need).
We are not interested in connecting to the internet from the SCO box (yet),
so any DNS/DHCP/LMHOSTS issues are local only.
File shares to a couple directories (for documents, images, etc.) on the SCO
box are made possible through VisionFS 3.0.919.
In looking further, the named daemon is running (so how do I kill it?),
along with a hell-spawned army of other demonic assistants.
I'm far too inexperienced to give you more than that without specific
'how-do-I-do-this' steps.
The problem essentially manifests when trying to run an executable that is
stored on the file server from a windows workstation. At this point, we
only have two such applications: VisionFS Profile Manager, and our ODBC
link. Given these two products are produced by different vendors for
different purposes, I am assuming the problem lies in the SCO configuration,
not with the products.
I have added the server IP and name, and one of the clients IP and name (my
machine) to the /etc/hosts file. There is also an entry for localhost in
there.
I have created the C:\WINDOWS\LMHOSTS file as directed, but the lag still
remains. Network setting in windows include the DNS settings pointing to
the IP address and name of the server.
I really don't care about the lag in VisionFS Profile Manager - I'm the only
one using it. But the ODBC lag is a problem. ODBC times out continually
because of the lag, which makes the whole thing damn near useless. If I can
get it going once, all future connections from that workstation are good,
but a reboot kills the link, and the other workstations still have trouble
connecting that first time.
While I'm at it, is there a definitive FAQ or list of "essential" processes
(daemons?) so I can wipe some of the superfluous crap sitting on this
machine? What are the bare essentials for a secure, fast, file server?
What processes are a must, and which are optional? All I need is local
access to files and directories within a single office. Fast, clean, and
secure.
Jim Richardson <j...@dnrcrd.dnr.state.ga.us> wrote in message
news:7of0sh$cgc$1...@nnrp1.deja.com...
> In article <7oep3l$2j...@enews1.newsguy.com>,
> "Waif" <wa...@ods.ods.net> wrote:
> > Greetings,
> >
> > New to Unix, so pardon any idiocy. I have been playing with a SCO
> 5.0.5 OE
> > server for a couple months now, as time permits (which ain't often).
> >
> > I am presently trying to clear up the following issue: The first time
> I try
> > to run a program off the SCO server via Windows95, the system takes
> 5-8
> > minutes to pull up the application, but subsequent attempts to run
> these
> > applications results in immediate access. I've tried this with
> VisionFS
> > Profile Manager (3.0) and with an ODBC link to our new manufacturing
> > software. Note that telnet and access to VisionFS shares through
> Network
> > Neighborhood does not appear to have this problem - both connect
> > immediately.
> >
> > I posted previously and was told this was more than likely an issue
> with DNS
> > being improperly set up, which is highly believable.
> >
> > So now I'm looking for resources on how to set it up, or more
> importantly,
> > advice on which type to set up. We have 1 SCO server (128MB Ram,
> Pentium
> > II 266, 6GB disk space). We are a small company with 25 desktop
> systems and
> > a single server. I can't justify spending the cash for another SCO
> box just
> > to act as a DNS server, so I want to run DNS off the one box we have
> now.
> > But I don't want to slow down this box (file server) too much either.
> I
> > assume I want to set up this box as a primary DNS server, but am
> concerned
> > that it will slow things down too much. Do I want a cache-only server
> > instead? Or maybe I don't need DNS at all??
> >
> > Also, is there a way to check if components are already set up so I
> can
> > strip them and start over?
> >
Jim Richardson
This is probably part of your problem. Since you really aren't using
DNS right now, you are adding much lag by calling it. Chances are I was
the one who recommended you set that up and if so, we'll Mea Culpa. For
now, get rid of DNS in the pcs.
Check and make sure you are looking at hosts on the Unix box. You should
have all the pcs listed in there with their names as set up in the 'PCS'
network config. Then set up LMHOSTS as recommended in the post from
Tony.
What is probably happening is that your pc is first trying to query DNS
on the Unix box which, like Sargent Schultz "Knows nothing". Once it
times out it looks in its pocket and finds it had the answer all along.
For now, don't kill of any of the daemons. That way can lead madness.
Also check out www.aplawrence.com for some great pointers.
Tony Lawrence
: More information it is.
: SCO was installed by an outside vendor from whom we purchased a new
: manufacturing software package. They know enough to get things running, but
: aren't wizards by any means.
Then the first thing you need to check is whether they knew enough to install
the proper patches: see http://www.aplawrence.com/newtosco.html for pointers.
: While I'm at it, is there a definitive FAQ or list of "essential" processes
: (daemons?) so I can wipe some of the superfluous crap sitting on this
: machine? What are the bare essentials for a secure, fast, file server?
: What processes are a must, and which are optional? All I need is local
: access to files and directories within a single office. Fast, clean, and
: secure.
http://www.aplawrence.com/Security/ might help you get started- there is an
awful lot you probably do not need to be running, especially since you aren't
connected to the greater world.
--
Tony Lawrence (to...@aplawrence.com)
SCO Articles, Tips, Book Reviews: http://www.aplawrence.com
Simon Hobson
>I posted previously and was told this was more than likely an issue with DNS
>being improperly set up, which is highly believable.
From memory (and I stand to be corrected), the default timeout for DNS
queries is about 90s, so if your PCs are trying DNS queries and failing it
would appear that they (or your SCO box) are trying 4 times.
Literally as I typed this I had a thought ...
The default configuration for SCO DNS servers includes a list of root
servers for the internet (with comments to change these to current values
as they could be wrong). If your system is not connected to the net, and is
mis-configured, it could be trying to contact some other server to get an
address.
In all the stuff below, where I have used <yourdomain.com> you should use
your own domain name (without the < & >). If you do not have a domain name
registered, then you can effectively use whatever you like as you are not
connected to the outside world - but if you use a non registered name and
later connect then your are likely to have problems.
As a starting point, look at your /etc/named.conf file. You should have
entries like :
zone "<yourdomain.com>" {
type master;
file "hosts";
};
zone "nnn.nnn.nnn.in-addr.arpa" {
type master;
file "rev";
};
These tell named about your local network.
At the bottom, you will probably have an entry like :
zone "." {
type hint;
file "root.cache";
};
This tells named what servers to go to if asked for an address NOT in
<yourdomain.com>.
You need to be logged in as root for the following steps, so be VERY
careful.
Check the file root.cache (in /etc/named.d by default) and comment out any
servers listed (putting a ";" at the front of every line should do it).
Then tell named to reload it's database :
Get the process ID of named, either by "cat /etc/named.pid" or by "ps
-ef|grep named". Tell named to reload it's database by "kill -HUP <pid>"
where <pid> is the number found in step 1.
If this is the problem, when first connecting from the PCs, one of three
things will happen :
1) The delay will be gone
2) You will get an error message about not being able to resolve a host
name
3) There will be no change, in which case I don't think it is likely to be
a DNS problem.
On the PCs, chack that in the DNS tab of TCP/IP properties :
A hostname is entered in the hostname box.
<yourdomain.com> is entered in the domain box.
Only the IP address of the unix system is listed under servers.
<yourdomain.com> is listed in the search domains box.
As far as system loading is concerned, I wouldn't worry. Our network makes
extensive use of DNS (around 140 nodes over 4 sites, all using one DNS
server), and our DNS server has to handle our internet access as well.
Quickly checking I see that in nearly 8 days uptime, the DNS server has
consumed a total of under 3 1/2 hours of CPU - this works out (allowing for
quiet times etc) as only a couple of percent of our capcity. On a small
network like yours I doubt if you will see anything like that loading. We
have an Acer server with dual Pentium II 200's, 1G ram, and about 56G disk.
Hope this is some help, Simon
Kenneth McCormick
[snip]
|We are not interested in connecting to the internet from the SCO box (yet),
|so any DNS/DHCP/LMHOSTS issues are local only.
|
|In looking further, the named daemon is running (so how do I kill it?),
|along with a hell-spawned army of other demonic assistants.
|
|I'm far too inexperienced to give you more than that without specific
|'how-do-I-do-this' steps.
You've done well so far Waif.
1) Do a ps -ef | grep named | grep -v grep and find the pid of named and kill it.
Now named (dns) is turned off.
2) Rename the file /etc/named.boot to /etc/named.boot.orig. Now named will not
start again after a reboot. DNS is now permanantly off.
3) Created entries in your unix /etc/hosts and your windows C:\WINDOWS\HOSTS
for tcp/ip that look like this:
127.0.0.1 localhost
10.1.1.1 spock.mydomain.com spock
10.1.1.2 kirk.mydomain.com kirk
10.1.1.3 sulu.mydomain.com sulu
4) Create entries in your unix /etc/lmhosts and your windows C:\WINDOWS\LMHOSTS
for Netbios (vision) that look like this
10.1.1.1 spock
10.1.1.2 kirk
10.1.1.3 sulu
You can create these files once, copy them to /tmp and copy them to each
computer if /tmp is a vision share. As others mentioned, you can get by without
dns, and just use netbios and /etc/hosts. Netbios traffic is minimized in this
case becasue of the duplicate lmhosts files on each comp. Dns lookups can
really slow a network.
In a private network like yours, be sure to choose valid ip addys like
the ones I listed above. Search this ng on dejanews for that topic.
As far as necessary daemons, you don't need fingerd, talkd, nis, dns,
nfs, or snmp. Others can comment on gated and routed.
Your system will be very snappy with sufficient RAM, L2 cache, and cpu speed.
Let us know how it's going :)
Ken
Waif
First, thanks to each of you for the help so far.
At this point, I have slain the named daemon, and have removed the
/etc/named.conf file to prevent a reversal of the exorcism.
The 'ps -ef | grep named' command shows no processes, so it's good and gone.
I have set up an /etc/lmhosts file as follows (I'm only trying to get this
to run from 1 client initially):
192.1.1.1 nameserver
192.1.1.2 client1
(The names have been changed to protect the innocent.)
The C:\WINDOWS\LMHOSTS file matches 100%
I have an /etc/hosts file as follows:
127.0.0.1 localhost
192.1.1.1 nameserver.domain.com nameserver
192.1.1.2 client1.domain.com client1
Again, the C:\WINDOWS\HOSTS file is identical.
As for the lag - the system is down to waiting only 50-60 seconds before
things pop up, so it's moving closer.
Any other tips?
Kenneth McCormick <mrsm...@hotbot.com> wrote in message
news:MPG.1217b7523...@news.earthlink.net...
> Recentky, wa...@ods.ods.net said...
>
> [snip]
>
> |We are not interested in connecting to the internet from the SCO box
(yet),
> |so any DNS/DHCP/LMHOSTS issues are local only.
> |
> |In looking further, the named daemon is running (so how do I kill it?),
> |along with a hell-spawned army of other demonic assistants.
> |
> |I'm far too inexperienced to give you more than that without specific
> |'how-do-I-do-this' steps.
>
Jim Richardson
"Waif" <wa...@ods.ods.net> wrote:
> The saga continues...
>
> First, thanks to each of you for the help so far.
>
> At this point, I have slain the named daemon, and have removed the
> /etc/named.conf file to prevent a reversal of the exorcism.
> The 'ps -ef | grep named' command shows no processes, so it's good and
gone.
>
> I have set up an /etc/lmhosts file as follows (I'm only trying to get
this
> to run from 1 client initially):
>
> 192.1.1.1 nameserver
> 192.1.1.2 client1
>
> (The names have been changed to protect the innocent.)
>
> The C:\WINDOWS\LMHOSTS file matches 100%
>
> I have an /etc/hosts file as follows:
>
> 127.0.0.1 localhost
> 192.1.1.1 nameserver.domain.com nameserver
> 192.1.1.2 client1.domain.com client1
>
> Again, the C:\WINDOWS\HOSTS file is identical.
>
> As for the lag - the system is down to waiting only 50-60 seconds
before
> things pop up, so it's moving closer.
>
> Any other tips?
Make sure that on the windows box the only protocol enabled is tcp. If
ipx or netbeui are enabled kill them off.
>
> Kenneth McCormick <mrsm...@hotbot.com> wrote in message
> news:MPG.1217b7523...@news.earthlink.net...
> > Recentky, wa...@ods.ods.net said...
> >
> > [snip]
Kenneth McCormick
|In article <7on1v6$1j...@enews2.newsguy.com>,
| "Waif" <wa...@ods.ods.net> wrote:
|> The saga continues...
|> First, thanks to each of you for the help so far.
Cheers.
|> At this point, I have slain the named daemon, and have removed the
|> /etc/named.conf file to prevent a reversal of the exorcism.
[...]
|>
|> As for the lag - the system is down to waiting only 50-60 seconds
|before
|> things pop up, so it's moving closer.
|>
|> Any other tips?
|
|Make sure that on the windows box the only protocol enabled is tcp. If
|ipx or netbeui are enabled kill them off.
|
|Jim Richardson
|I like NT because it constantly reminds me of my daughter.
|"Honest Daddy, I wasn't doing anything and it just broke."
And then you might take a look at what the output of
ping -f client1
netstat -i
is after to see if you have a cable problem or some other act of weirdness.
Simon gave a lot of good information it would behoove someone who is attempting
dns to work through it. "DNS and Bind" is an excellent reference also.
It's really nice to hear from someone who has dns running so smoothly.
OS 5.0.x seems to ship with a nasty set of default named
files that are examples of a dns setup in a company with sco in the
domain and we've all seen that it runs right after ISL, thus thrashing domain
name space for any poor chap who happens to be connected to a valid domain
or who gets curious and tries ppp to their isp, OR who tries visionfs.
Then there are the incessent lookups to A.ROOT-SERVERS.NET and
B.ROOT-SERVERS.NET and C.ROOT... with all those requests from
broken thrashing sco domain lookups, oh man. They must just
scratch their heads over there when they look at their log files.
Ken
Waif
I picked up "DNS and BIND - 3rd Edition" last night and began browsing
through it. I was hoping to get around DNS temporarily (it's a very small
private network), and attack that dog later.
Protocols in Windows are TCP/IP and IPX (we are still running via NetWare
3.12 until the Unix box is completely set up). Browse Master is disabled
and LM Announce is set to "No"
"ping -f client1" returns: "PING client1.domain.com (192.1.1.2): 56 data
bytes", then spits out dots until I stop the ping process.
"netstat -i" returns:
Name Mtu Network Address Ipkts
Ierrs Opkts Oerrs Coll
net1 1500 domain nameserver 188213 0
190269 0 20277
lo0 8232 loopback localhost 7841 0
7841 0 0
atl0* 8232 none none No
Statistics Available
As for what that all means... I assume the zeros in the errs columns are
good, but am somewhat worried about the collisions (should they be that
high?). atl0 I think is the SCSI card.
Cable problems shouldn't be an issue. Telnet has been running without a
pause for six or seven days now. Fires very nicely, and access to VisionFS
shares is instant. It's just when trying to execute something for the first
time that the lag appears.
I can try to get actual died-in-the-wool, tru-blue DNS up and running
following previous posts (which is where the network would have ended up
eventually anyways). Are there any steps I should take to clean away what
I've done so far (e.g. kill the /etc/hosts and /etc/lmhosts files)?
- James
Kenneth McCormick <mrsm...@hotbot.com> wrote in message
news:MPG.121959867...@news.earthlink.net...
> Recentky, j...@dnrcrd.dnr.state.ga.us said...
> |In article <7on1v6$1j...@enews2.newsguy.com>,
> | "Waif" <wa...@ods.ods.net> wrote:
> |> The saga continues...
> |> First, thanks to each of you for the help so far.
>
> Cheers.
>
> |> At this point, I have slain the named daemon, and have removed the
> |> /etc/named.conf file to prevent a reversal of the exorcism.
>
> [...]
>
> |>
> |> As for the lag - the system is down to waiting only 50-60 seconds
> |before
> |> things pop up, so it's moving closer.
> |>
> |> Any other tips?
> |
> |Make sure that on the windows box the only protocol enabled is tcp. If
> |ipx or netbeui are enabled kill them off.
> |
> |Jim Richardson
> |I like NT because it constantly reminds me of my daughter.
> |"Honest Daddy, I wasn't doing anything and it just broke."
>