This looks promising, I'll give it a whirl -
Joe
Henrik Stoerner wrote:
Hi Joe,
On Thu, Nov 01, 2007 at 03:20:12PM -0700, Sloan wrote:
So, the $64 question: Is there anything in hobbit, or on the horizon, which will allow hobbit to serve as a drop-in replacement for bb, including the failover capability?
The BB "failover" script does two things: It makes the network tests run on the failover server if the primary BBNET server cannot be ping'ed; and it enables alerts being sent from the failover server if there is no connection from the failover server to the primary BBPAGER server.
The network-test failover is fairly simple to do. I've attached two scripts here, both of which must run on the backup/standby/failover server:
failover.sh - goes in ~hobbit/server/ext/ Add a section to hobbitlaunch.cfg with
[failovercheck] ENVFILE /usr/lib/hobbit/server/etc/hobbitserver.cfg NEEDS hobbitd CMD $BBHOME/ext/failover.sh 10.0.0.1 hobbitnet.mydom.com
"10.0.0.1" is the IP of your primary Hobbit server, "hobbitnet.mydom.com" is the hostname (in the bb-hosts file) of the primary network test machine.
What this does is that it queries the primary Hobbit server for how long ago the network tests were updated. If more than 7 minutes ago it deems the primary network test node to be DOWN, and flags this via the file $BBTMP/primarynetDOWN. If the network test update was less than 7 minutes ago, it removes the file.
This is then used by the other script, which replaces the CMD in the "[bbnet]" section in hobbitlaunch.cfg.
failovernet.sh - goes in ~hobbit/server/ext/ When this runs to do the normal network tests, it will check for the presence of the $BBTMP/primarynetDOWN file. If this file exists, it picks up the IP of the primary Hobbit server from the file, and modifies the settings to report data to both the normal (local) Hobbit server, and to the primary server. If the file does not exist, it will just run the network tests the normal way. So to run this, modify the [bbnet] section in hobbitlaunch.cfg and change the CMD setting to "$BBHOME/server/ext/failovernet.sh"
The alert failover is different, because Hobbit doesn't have a separate BBPAGER server - alerts are sent from the same host that handles the Hobbit data collection and webpages. A solution to this has been implemented for the next release, where the alerting module can be distributed onto multiple servers, but only one of them will send alerts at any given time.
Regards, Henrik
To unsubscribe from the hobbit list, send an e-mail to hobbit-unsubscribe at hswn.dk