I am even more sure now By what i see in the graphs you should change Max clients to 30 for now.
Regards Luchy ---------------------------------------------------------------- Email : [EMAIL PROTECTED] ----- Original Message ----- From: "Ira Abramov" <[EMAIL PROTECTED]> To: "IGLU Mailing list" <[email protected]> Sent: Tuesday, January 18, 2005 7:55 PM Subject: Re: new faster CPU = less open connections?! > Quoting Ira Abramov, from the post of Sun, 16 Jan: > > Quoting Shachar Shemesh, from the post of Sat, 08 Jan: > > > > > Either someone changed the apache config to keep less open sockets or > > > less concurrent processes, or apache, of it's own accord, came to the > > > conclusion it can make do with less. You tell us which. > > > > I didn't really think Apache can make its own decisions like that, but > > than I took a long look at this week's graph for "number of apache > > procs" right here: > > http://jenna.scso.com/hotsanic/apps/apache2.html > > since thursday morning it seems that for no apperent reason (or logic) > > Apache has changed it's minservers/maxservers bahavior (though no new > > binary was installed and I haven't adited the config file in 10 days > > (it's dated 6 January). loads on the site are the same, as are the CPU > > and netstat connections: > > doesn't anybody have any ideas? > > the config for that was not moved from Debian's defaults, which is the > prefork MPM and the settings are: > StartServers 5 > MinSpareServers 5 > MaxSpareServers 10 > MaxClients 20 > > as for my other problem: > > > http://jenna.scso.com/hotsanic/netstat/connections.html > > The large number of "syn recieved" since my move to Bezeq... they claim > they have no delays (as Shachar suggested) but this graph is still odd, > since I'd expect a delayed handshake will be followed by a delayed FIN > sequence as well, but it is not the case here. Could this be a syn > attack of some sort? how can I tell for sure? should I turn on some sort > of syn cookie mechanism? Right now many of the web users complain about > coming-and-going slowness that we can't explain or measure (it COULD be > a problem with the link only to specific ISPs). > > Also, the NOC at Nezeq Benleumi say this could be blamed on a NIC with > "a problematic buffer". I am not sure what they mean but since I have > two NICs on that board (an e100 and an e1000) I don't mind switching to > test. > > any ideas welcome, as always.... > > > -- > Not an expert > Ira Abramov > http://ira.abramov.org/email/ > > ================================================================= > To unsubscribe, send mail to [EMAIL PROTECTED] with > the word "unsubscribe" in the message body, e.g., run the command > echo unsubscribe | mail [EMAIL PROTECTED] > ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
