Another option is to tie it to the number of file descriptors available to your process and cap it to some value slightly below that. If the user specified a value above the cap then emit a warning and use the lower value.
-Jason Martin > -----Original Message----- > From: > [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] > org] On Behalf Of Martin, Jason H > Sent: Tuesday, October 18, 2005 11:52 AM > To: Mark Burgess; David Masterson > Cc: [email protected] > Subject: RE: Max cfservd connections > > > Even with a fully loaded Sun F15K with several GigE > interfaces and some really fast storage? > > OK, I wish I had that, but there are all kinds of > possibilities :> At a minimum the cap should be documented, > but preferably removed entirely. > > -Jason Martin > > > -----Original Message----- > > From: Mark Burgess [mailto:[EMAIL PROTECTED] > > Sent: Tuesday, October 18, 2005 11:46 AM > > To: David Masterson > > Cc: Martin, Jason H; [email protected] > > Subject: RE: Max cfservd connections > > > > > > > > Anyone with such a system would be out of their mind to > > expect a single machine to cope with such a load, > > > > M > > > > On Tue, 2005-10-18 at 10:52 -0700, David Masterson wrote: > > > A large system with many network connections in an environment > > > supporting thousands of machines and (for whatever reason) > > no use of > > > SplayTime? > > > > > > Also consider the reverse -- a tightly controlled *small* > > system that > > > a user wants to severely limit the maxprocesses to prevent > > runaways, > > > so he lowers maxprocesses. > > > > > > Mark Burgess wrote: > > > > Fair enough - it was meant to be a silly number. I did not > > > > anticipate anyone contemplating this. I would be > > interested to know > > > > he circumstances in which it is actual to expect 1000 > simulaneous > > > > connections to a single machine. > > > > > > > > M > > > > > > > > On Tue, 2005-10-18 at 10:13 -0700, David Masterson wrote: > > > >> Arbitrary numbers should be configuration items. ;-) > > > >> > > > >> Mark Burgess wrote: > > > >>> This is just an arbitrary number. Most systems will not > > want to go > > > >>> higher -- but if you do, then change it. M > > > >>> > > > >>> On Mon, 2005-10-17 at 16:14 -0700, Martin, Jason H wrote: > > > >>>> I'm curious, does anyone know why cfservd is limited to 1000 > > > >>>> concurrent connections? > > > >>>> > > > >>>>> From the latest cfservd.c: > > > >>>> if ((CFD_MAXPROCESSES < 1) || (CFD_MAXPROCESSES > > > 1000)) { > > > >>>> FatalError("cfservd MaxConnections with silly > > value"); > > > >>>> } > > > >>>> > > > >>>> Thank you, > > > >>>> -Jason Martin > > > > > > > > > > _______________________________________________ > Help-cfengine mailing list > [email protected] > http://lists.gnu.org/mailman/listinfo/help-> cfengine > _______________________________________________ Help-cfengine mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-cfengine
