bzfs problem

[swine]

I'm running bzfs on linux mandrake. P2 400. All is well for a while then I can see by my cable modem lights that something weird is going on. Checking the system I find that bzfs is now using 95% of my cpu. (it was using .3%) I log on to the server and there are only three or four people there. Lagcheck pegs them at 3 to 4 secs lag, then boom, everything goes back to normal. But by this time everyone has left.
Am I doing something wrong?
Is there some bz god that I must sacrifice a newbie too?
Is my system too obsolete?
Can someone send me a new computer?
Is it Beer time?
inquiring minds want to know!

[Fiberchunks]

iirc, I was a post (I think on sourceforge), about bzfs slowing down after x many minutes, and eating cpu cycles...hrm - where was that?

Heh, that was a post by dutchrai on the other forum (ezboard) about a windows server borking and eating all the cpu...

Hrm, I didn't think that affected the *nix version of bzfs.

Couple of questions:
1. what does the output of top show (%cpu usage and %mem consumed)
2. Do you have any other rogues processes floating around?
3. what does your bzfs commandline look like (post it)

We'll try and get it straightened out

Peace

[swine]

Here is the init script I use. Its a combination of the one generated when starting a random server (as read by "treeps") and bits of my other scripts.

bzfs -pf -r -h -mp 20 -lagdrop 5 -lagwarn 1000 -maxidle 300 -passwd "" -public "Swine's Sty of DOOOM {random acts}" -publicaddr "24.80.130.30" -ms 10 -t +r -j +s 40 -sa -st 5 -sw 1 -ttl 255
[edit]removed url[/edit]
Thanks eh!

[Fiberchunks]

Here is the init script I use. Its a combination of the one generated when starting a random server (as read by "treeps") and bits of my other scripts.

bzfs -pf -r -h -mp 20 -lagdrop 5 -lagwarn 1000 -maxidle 300 -passwd "" -public "Swine's Sty of DOOOM {random acts}" -publicaddr "24.80.130.30" -ms 10 -t +r -j +s 40 -sa -st 5 -sw 1 -ttl 255
Thanks eh!

the link is busted -- pm me with it.

what is -pf? I know that -p specifies a port number (unless you use the default -- but I am unaware of pf... as a matter of fact -pf is an invalid entry -- you could either use -pr <port>, or -p <port> -- so is that a typo?

I would also add (even though I'm not absolutely positive that it's necessary anymore) -requireudp to your command line.

Those are the only things I can think of off the top of my head -- try modifying that commandline with my recommendations and let me know how it goes...

By the way - I deleted that link, as it doesn't work, and you wanted it gone anyway

Peace

[Fiberchunks]

The only thing I can suggest (as bzfs does look to be doing something odd...is that you manually start the server from the cli ...

Try this:
bzfs -p 5155 -r -j -public "blah" -publicaddr "xxx.xxx.xxx.xxx:5155" -mp x,x,x,x,x -ms x -ttl 255 -requireudp <etc.>

Also note that cable modems that a history of being flaky (since it's a shared node connection, affected negatively every time someone connects to the network on your node.)

Also, what version of the server are you running, the latest (1.7gX).... and really important, what ports do you have open on any routers / firewalls you're running --- note that these should be open (at the bare minimum):

5155 TCP
17200-172xx UDP (where xx is max players +a couple) , for example if -mp 10, then 17200-17212 should be open...

Now, if you use any port besides the default (5155), then note that in your command line, you MUST change 2 things; the parameter to -p, and the port in the -publicaddr part (e.g. if you want to use 7500, then -p 7500 -publicaddr "xxx.xxx.xxx.xxx:7500" ).

Try that out -- and be sure that your command line is correct, and includes only the commands that are in the man page...

Try these out, and let me know how it goes.

Peace

[Chestal]

I'm running bzfs on linux mandrake. P2 400. All is well for a while then I can see by my cable modem lights that something weird is going on. Checking the system I find that bzfs is now using 95% of my cpu.

I have seen something similar on ducati sometimes. The server eats up all available CPU instead of behaving normally. It doesn't really affect gameplay on ducati. I haven't checked whether it is sending more messages or something when doing so, ducati's connection to the outside is only limited by the 34MBit/s line of the university, so maybe that's the difference. There sure is some bug there relating to a player slot which is in an undefined state or something. I only saw it rarely now and it kept healing itself, soI didn't really tried to pursue this issue.