Re: [Spong-users] Spong for a High Availability cluster
Brought to you by:
apremselaar,
sljohnson
|
From: Stephen L J. <sjo...@mo...> - 2001-02-20 21:15:51
|
> Hi all!
> I'm using Spong in a cluster of three nodes. I want to check the network
> services every 180 seconds so I put the the SLEEP variable to 180. But as I
> have Spong on every server each one check the servers listed at spong.hosts
> every 180 sec. so I can't control that they do the check at the same time.
You will not be able to keep the spong-network's in sync. The Spong programs
are designed to stay out of sync. They will sleep the amount you specify +/-
10%. Imagine 200+ servers booting up some type of outage. All 200
spong-clients trying to send status messages to the spong-server at once. Poor
little spong-server... :)
>
> I have the three servers in the variable SPONGSERVER for the three nodes, so
> each one is server of each other, I tried to put in spong.hosts just one
> server -the localhost- and in spong.groups the three servers, but after
> doing it the server was receiving client-data just of itself and when
> viewing "groups" in the browser the others were in purple.
The purple servers are old, stale data. They haven't been updated in 3 x
$SPONGSLEEP{'spong-network'} seconds. Check your $SPONGTMP/spong-network.log
or syslog log files for connectivity errors all of the servers.
>
> How can I sinchronize the checks for the cluster?
Even if you eliminate the randomness from the SLEEP code. If will be had to
keep the spong-network programs in sync. A check may fail on the first try, a
server may have a little bit longer response time than the other servers, etc.
>
> I would like receive any advice for using Spong on a high availability
> cluster.
|