Re: Failover clustering of storage servers

On Thursday 23 September 2004 16:03, Dario Lopez-Kästen wrote:

> The slave runs a heartbeat-script (essentially a ZEO-client script) that 
>   chekcs if the master is running. If not it shuts down the master and 
> reconfigures itself to take over the masters identity.

.....and starts the ZEO server on the client. Note that you have to keep the 
ZEO on the client shut down during normal operation to allow replication to 
work.

> At most we loose aprox. one minutes data.

Do you plan for your standby to be read-only? thats easier.

If it is read-write then you will need some way of updating the master with 
changes made on the secondary. Note that you cant use normal DirectoryStorage 
replication for this because it doesnt have any way of resolving the. 
conflicting changes made on the master during that lost minute.

(thinking aloud..... the replication tool could be enhanced to efficiently 
back out any changes made in that lost minute. I can discuss implementation 
strategies if you fancy tackling this)

> Does this sound like something that can be done with directory storage? 

sure

> Has anybody done something similar to this?

Ive experimented with it, but so far all my production replicas are cold 
standbys

-- 
Toby Dickenson


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php