Your options are fairly limited in this regard.
Traceroute and PING are fine, but unfortunately they
do not use the same transport protocol as Half Life.
HLDS uses UDP as its primary transport for data. UDP,
from a networking point of view, is not as important
as TCP or ICMP.
You didn't mention it in your email, but if you
connect to your servers on the local network i assume
the connection is fine and there is virtually no
packet loss. The packet loss is being experienced by
users who connect via the Internet?
If this is the case all you can do is either identify
the problem locally and then work with your network
admin to resolve it...otherwise you are screwed.
Here is how I would initially approach this. First
thing I want to do is call whoever is incharge of your
datacom gear <are you at work?> find out if there has
been any equipment installed recently that would
account for this. One product that comes to mind is
Floodgate. These products are used as "protocol
shapers" they prioritize traffic based on virtually
any metric you desire. Some router and switch
software is also capable of performing a limited
amount of this. If this is the case perhaps you can
convince them to up the priority of UDP traffic going
to and from your servers.
Assuming that comes up negative I would then want to
find out if the problem is on my local segement. UDP
is connectionless and if packets don't make it to
their destination it is no big deal. TCP, on the
other hand requires receipt for each packet. On your
Linux box learn how to use TCPDUMP. I cannot even
begin to explain how to use it. It is very powerful
and will be able to give you answers. If you have a
Sniffer, thats even better. If you do not understand
protocol analysis you are either screwed or you have
just found a great way to learn about it. You'll
never be out of work if you can get good at it.
If you have a layer 1 problem on your segemnt it will
show up in the form of CRC errors, typically. Bad
cable, bad media adapter whatever. You might find you
have a bad nic or even worse, your switch is getting
ready to crap out. If you can be reasonably certain
that TCP packets are all getting where they are
destined without to many retransmissions, then you can
be fairly certain that the software and hardware on
the computer are fine. You will need TCPDUMP or a
sniffer to prove this, though. Simply connecting to a
web site isn't good enough.
If you are pretty sure that there isn't a physical
problem with your segment, then you can now get ready
for the fun. You will have to find a user of your
servers who is experiencing the problem and is willing
to work with you while you do the testing. You'll
have to setup a server <a new server...not one of the
ones you have now> so that only he is connected to it.
He has to have equipment on his end that will let him
count <and view> packets basically you are going to
let him connect and play for like 30 seconds. Look at
your logs. If you received 60 packets from him, he
should have sent you 60 or at least no more than 70 or
so. If you sent him 70 he shouldn't have received
fewer than 60 - 65.
>From this point things get pretty complex. If the
packets aren't being dropped on your network, you
probably have very little control over the solution
anyway.
Having written all this, the short answer to your
question is TCPDUMP or Sniffer if you want to start
looking into problems with packet loss.
Hope this gives you some ideas.
--- David <[EMAIL PROTECTED]> wrote:
>
> Need a little help here, as i don't know where to
> look next....
>
> Question: what tools can i use to investigate packet
> loss problems?
>
> I have 3 servers running on two machines, set up on
> the lan at work, but
> open to the net. and all servers are getting bad
> packet loss troubles.
> This problem started suddenly, after over 8 months
> of everything fine
> (prob started before the new patch, so i don't
> believe it's related).
> and since both machines have the problem, i'm
> guessing its a local
> network problem... Pings are still fine for
> everyone, but the connection
> time as increased (often times out), and the packet
> loss is way high. I
> have tried running the servers clean, without
> metamod & all, but the
> problem is the same. I have no other service on the
> machines other than
> hlds_l, and i'm the only user...
>
> anyone have an idea what could be causing these
> problems? and how could
> i investigate farther?
>
> thx
>
>
__________________________________________________
Do You Yahoo!?
Get email alerts & NEW webcam video instant messaging with Yahoo! Messenger.
http://im.yahoo.com
