LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: [lvs-users] in ldirectord.cf, the meaning of 'checkcount' is ambiguo

To: "Robinson, Eric" <eric.robinson@xxxxxxxxx>
Subject: Re: [lvs-users] in ldirectord.cf, the meaning of 'checkcount' is ambiguous
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
From: Simon Horman <horms@xxxxxxxxxxxx>
Date: Thu, 14 Jul 2011 18:18:03 +0900
On Thu, Jul 14, 2011 at 01:33:22AM -0700, Robinson, Eric wrote:
> > -----Original Message-----
> > From: Simon Horman [mailto:horms@xxxxxxxxxxxx]
> > Sent: Tuesday, July 12, 2011 7:56 PM
> > To: Robinson, Eric
> > Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> > Subject: Re: [lvs-users] in ldirectord.cf, the meaning of 'checkcount'
> is
> > ambiguous
> > 
> > On Fri, Jul 08, 2011 at 04:45:17AM -0700, Robinson, Eric wrote:
> > > The ldirectord man page states that 'checkcount' only works with
> ping
> > > checks. Then the next sentence goes on to give an example of
> checkcount
> > > being used in conjunction with connectimeout/negotiatetimeout,
> neither
> > > of which are related to ping checks. Can someone please clarify?
> Horms?
> > 
> > Hi Eric,
> > 
> > sorry for not responding earlier.
> > 
> > I've taken a look at the code (in git) and to be honest I don't
> > think that this is my handiwork. But none the less I believe that
> > the situation is:
> > 
> > * checkcount only applies to ping.
> >   There is a loop in the ping check that will try
> >   ping up to checkcount times, breaking out of the loop
> >   if a ping is successful
> > * checkcount is deprecated in favour of failurecount
> > * failurecount works similarly to checkcount, except that it
> >   it operates at a higher level and applies to all types of checks
> > * It appears that specifying both failurecount and checkcount would
> have
> >   a multiplier effect, though only for ping checks
> > 
> > I will fix up the sample ldirectord.cf to make use of failurecount
> > instead of checkcount.
> 
> That's delightful, Horms, thanks for the clarification. I don't think my
> version of ldirectord has the failurecount option. It's been in
> production since 2006 and is currently managing about 900 virtual
> services. I've never wanted to mess with it much, considering that it
> has worked very well for us. The failurecount option may be adequate
> motivation.

Hi Eric,

before upgrading I would check the source code to see if
checkcount (or some other option) meets your requirements.

_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/

LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users

<Prev in Thread] Current Thread [Next in Thread>