Real gets full loa after crash and reinsert.
janklopper at gmail.com
Sun Mar 12 10:37:54 GMT 2006
Strange enough, the servers all keep a weight of 1, all have about 30k
inative connections, and 3.4k active connections, but i can see that all
actual traffic (using mrtg), is going to, and commnig from the just rebooted
Il see if i can replicate the behaviour later this week, and make some
snapshots of the mrtg stats.
On 3/11/06, Joseph Mack NA3T <jmack at wm7d.net> wrote:
> On Sat, 11 Mar 2006, Jan Klopper wrote:
> > Joe,
> > Im using round robin, (as i posted),
> hadn't thought about that with the thundering herd problem.
> > but the number of ActiveConn and InActConn seems to be the
> > same on both the servers that stayed online, and on the
> > one taking all the hits.
> what's the consequence of this in the current problem?
> I thought you were getting all new connections to the new
> realserver (unless I wasn't reading carefully).
> > I triplechecked , but the other 3 servers didn't get a different weight.
> > I figured round-robin scheduling (which is all i need , since all nodes
> > equal) would never have this "thundering_herd" problem.
> I would have thought so too, now that you mention it.
> Joseph Mack NA3T EME(B,D), FM05lw North Carolina
> jmack (at) wm7d (dot) net - azimuthal equidistant map
> generator at http://www.wm7d.net/azproj.shtml
> Homepage http://www.austintek.com/ It's GNU/Linux!
> LinuxVirtualServer.org mailing list - lvs-users at LinuxVirtualServer.org
> Send requests to lvs-users-request at LinuxVirtualServer.org
> or go to http://www.in-addr.de/mailman/listinfo/lvs-users
Innerheight Internet Diensten
More information about the lvs-users