Flooding the IRCD with a simple copy & paste

[fluid]

Sometimes I find intresting news articles and I post them in my channel to share the story or whatever..

Sometimes if I paste too much text, my ircd server NETSPLITs.

Either that or, Ill loose 1/2 my users and they will quit via "Connection Reset by Peer" right after I paste the text.

Is this normal?

I also notice right after the netsplit, if I rehash one server, the users come right back.


Thanks.

[Matridom]

rehashing causes the IRC to reload the conf file, if your link blocks are set to "autoconnect" then it will cause the servers to re-sync. it may be better to /connect servername rather then reload the file.

when you say "large articals", how large?

What type of connection is your server using?
are you using an operblock?
what is your max send/receiv queue set to?

[fluid]

when you say "large articals", how large?

say between 15 to 35 lines of text.

-
What type of connection is your server using?

basic cable
-

are you using an operblock?

Yes
-

what is your max send/receiv queue set to?

class client
pingfreq 90;
maxclients 500;
sendq 100000;
recvq 8000;

class server
ping freq 90;
maxclients 10;
sendq 1000000;
recvq 8000;
connfreq 100;

I am using autoconnect in my Link block.

Thanks.

[Matridom]

I would increase the recvq of servers class by removing the line completly(do you really want the server to limit it's receive queue to the same as what a client can accept?). Secondly, be aware that as an ircop, you are exempt from fake lag and can *easily* flood the server.

IRcop's are supposed to be responsible enough to not do that...

[fluid]

yea my mistake. i read it wrong.

recvq servers 8000; was never there.
that line was removed by default conf.
and was never added.

i thought it was there i was wrong.

Yes, I guess wont paste storylines anymore.
Ill just post a link to the article instead.

[Syzop]

'Connection reset by peer' is not the ircds fault at least. And I regulary paste like 40 lines on my ircd (*NIX) ;p

Only things that could have happened and is ircd-related is a quit of "Max SendQ Exceeded" (though, your sendq of 100k sounds fine), but since you say that didn't happen... ;)

Perhaps something windows related. Windows for some reason does all kind of weird things (mass disconnects) during mempeaks or low memory, though I personally only get that when I have a lot (thousands) of clients during clonetests ;).

Btw, recvq is ignored for servers :).

[medice]

What type of connection is your server using?

basic cable

what't "basic cable"? which upstream does the machine have?