[prev in list] [next in list] [prev in thread] [next in thread] 

List:       keepalived-devel
Subject:    Re: [Keepalived-devel] Node state notification
From:       Lhunath <Lhunath () pandora ! be>
Date:       2006-03-21 11:05:53
Message-ID: op.s6rfr3k7ym63yo () local
[Download RAW message or body]

>> Now, what we're looking for is the following:
>> Say when node A goes offline, node B takes over. We want to run a  
>> script on node B when node A comes back online.
>> More generally speaking, we want to be able to run a script on the  
>> Master node when it detects a new node joining the keepalived group  
>> (coming online).
>
> If "on startup" is enough, one could think of an rc-script that connects  
> to the vip using ssh and executes whatever you want. But assuming  
> keepalived would also start on startup, this might take the vip from the  
> previous master and break that connection. So you might want to use  
> "preempt_delay" for this idea.
>
> Afaik the master does not "know" if anyone is listening to the vrrp  
> adverts it sends. So it could not recognise a machine coming back  
> online. At least this is how I understood.
>
>> I take it this will trigger an election, is there any way to run our  
>> script when an election occurs? We need to do some syncing of ourselves  
>> (need to keep a DB synced).
>
> This does not have to trigger an election. You could use "nopreempt" to  
> keep the master status on a lower prio machine. Maybe this helps you not  
> having to execute that script?

Currently we're doing fault detection on two interfaces, and sending the
VRRP adverts over a third.

The first mentioned two interfaces are the one VIPs are being provided
for, the third is the one that's being used for the syncing (of keepalived
& the database). It's a dedicated and therefore somewhat secure channel.

Since the database servers are only listening on the interface used for
syncing (and VRRP adverts), and that interface has no VIPs, a node that
just came back online has no way of knowing which of the already online
nodes is the current Master node, so it doesn't know where to synchronize
it's database from. This is why the Master with the up-to-date database
needs to push his database synchronization onto the new node that came
online, and therefore needs to know the IP of the node that came online on
the VRRP interface.

With what you just explained, I understand this is impossible with the
current implementation.
-- 
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd_______________________________________________
Keepalived-devel mailing list
Keepalived-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/keepalived-devel

[prev in list] [next in list] [prev in thread] [next in thread]