It's probably not a good idea to run Gnome and Nautilus on a server (especially with a slower CPU). You should download and install a lighter window manager (IceWM or Blackbox are good choices) because Gnome and Nautilus are notorious for being slow on mid-low end machines.

Linux will grab all available memory and store it in cache and buffers, so that's normal. Run "free -m" command and look only at the " -/+ buffers/cache:" line. The amount under the "used" header is really the amount of memory being used.

Run "service --status-all" as root to see all the system services and their status (running/stopped). You can also type "setup" and look at the system services via a curses based menu. In "setup", you can move the cursor to a service and hit the "F1" key to see a description of the service. You should stop ("service servicename stop") any services that are not necessary.

Speck

I checked the hard drive with hdparm right after the first bog down and though I don't recall the exact mb per second throughput I do remember that it was quite good.

One reason that I do not think the memory is just cached is because if I am in KDE or Gnome after a memory max out the mouse gets very jerky, the system responds very sluggishly and the screen saver starts looking .... well strange. The system monitor shows <4000kb free.

I have been testing the system without startx and then starting kde or gnome to check the system monitor.

Is there a way to do that (monitor the system) in text mode?

Thanks all for the comments and insight. I will try to look at the swap partition when I return to work.

For console monitoring tools, there are top (process, cpu usage, memory), uptime (system load average), free (memory), sysstat (sar and iostat programs - usually run via cron), iptraf (very good network monitor). The sysstat and iptraf programs will probably not be on your system (you can get them from freshmeat.net).

While you are at the console, run "free" and post the entire " -/+ buffers/cache:" line to this forum. This will help narrow down the problem.

Once again, Gnome and KDE are very resource intensive and probably shouldn't be used on a server. If you need a window manager, take a look at some of the smaller (memory and cpu usage) ones.


Speck

Back at work and checking the penguin...

I just started the linux box.

Currently my -/+ buffers/cache line reads:
used 22 free 258
For the "mem" line:
Total 281 used 41 free 239 shared 0 buffers 5 cached 13
from the swap line:
total 250 used 0 free 250

I have just started a constant ping from 5 other computers and will let that run for awhile. I may also copy a large file over the network.

I will report the free -m stats after that.

I am also installing the redhat 7.2 on another box w/ a pentium processor to compare systems.

After a constant pinging the memory stats using free -m stayed the same.

However, after copying a 200 mb file from the linux box and then back to the linux box free -m showed:

Mem
total 281 used 278 free 2 shared 0 buffers 5 cached 242

-/+ buffers/cache
used 29 free 251

swap
total 250 used 0 free 250

is it still slowing down

Is the hard drive IDE? or SCSI? if IDE is it running on ATA33/66/100? Is it the only device on the cable? if CD-rom is there, is it being accessed? if SCSI - SCSI2, U, UW. U2 SCSI3? when booting, what does the SCSI adapter report as the device speed?

Ummm..i only read halfway down the first page so forgive me if this has been pointed out later on.

We have people talking about cpu's, ram etc. WHAT ARE YOU TALKING about. do you know what you are talking about?

Why has no one asked this person the network speed yet? is it 10/100/1000 mbit?

if its the first one then you definately need to sort that out. if it is mixed, then this will prob cause speed issues as well.

I don´t think the problem is in the speed of the CPU or amout of RAM.

I have been running RH as a file server and web server on our network for more than 2 years. It's been ruuning without any problems and it's faster than NT server I used berfore.

Here is setup:

12 workstations WIN95/98

Server:
PentiumPRO 200MHz
128MB RAM
20GB SCSI HDD
The Xserver is running all the time and apache server is getting under greater presure as we port part of our clinet win applications to PHP/Postgres applications.

I would recommend to rewiew your cmb.conf file.

Thanks again for all of the input.

The system still bogs down. It is on a 10 mb network but again...the winnt box flies through the processes.

I have checked the hard drive and get very good throughput. It is ATA-100 7200 rpm IDE.

I will take a look at the cmb.conf file but really wouldn't know what to look for.

I misspelled cmb.conf. It should be smb.conf

Here is my setup (it may be helpful). It's not perfect, but works with no problem.

;*******************section global*****************
[global]
dos filetime resolution = yes
fake directory create times = yes
delete readonly = yes
blocking locks = yes
oplocks = yes
locking = yes
share modes = yes
ole locking compatibility = yes
netbios name = server
character set = ISO8859-2
client code page = 852
workgroup = workgroup
server string = Linux
printcap name = /etc/printcap
load printers = yes
log file = /var/log/samba/log.%m
max log size = 50
security = SHARE
log level = 1
read raw = yes
write raw = yes
max xmit = 65535
getwd cache = yes
lpq cache time = 30
socket options = TCP_NODELAY IPTOS_LOWDELAY
dns proxy = no
encrypt passwords = yes
password level = 0
preferred master = no
os level = 0
null passwords = yes
dead time = 0
debug level = 0
strict locking = False
strict sync = False
sync always = False
domain master = no
;

I had a server on a network with 25 or so users, some on 10MB hubs and some on 100MB hubs. While this may effect the speed at which the data moves through the network, it was still very much usable.

I would expect it to do better with the 10 MB hub because it puts less demand on the server.

There are a couple of other things that come to mind on a LAN...

All systems should have the same mtu.

the socket options in the smb.conf file

Hmmm.... I hadn't thought about the socket options.

Right now we have the socket options set at:

TCP_NODELAY S0_RCVBUF=8192 S0_SNDBUF=8192

Is there a problem with this or is there an option that would maximize the socket options?

you need to test it without the setting to see what it will be like.

you could try


socket options = IPTOS_LOWDELAY TCP_NODELAY

or maybe IPTOS_THROUGHPUT

you will have to experiment


check out ...

man setsockopt