Venue Failover

[Ivan R. Judson]

I've been thinking we have data stores broken out enough to run a 
single data store for a bunch of venue servers, but maybe that's not 
true yet. Is it?

I don't want a list of backups yet, becuase then we have to put in code 
to decide which backup they all go to, and I think this is a temporary 
thing anyhow. Note: intergroup might make this need go away because it 
makes the event channel reliable, right?

--Ivan

On Wednesday, August 6, 2003, at 02:15 PM, Thomas D. Uram wrote:

>
> A couple quick thoughts:
>
> - What about a list of backup servers?
> - SC Global has talked about remote datastores, which would allow 
> multiple venue servers to use the same datastore.  This backup 
> scenario could obviously benefit therefrom.
>
> Tom
>
>
> Ivan R. Judson wrote:
>
>> Here's the simple mechanism I was planning on inserting for venue 
>> failover,
>> I've followed that up with open issues and concerns. I'd like to get
>> feedback on how to close the open issues.
>>
>> 1. Add to the Venue Server config file: BackupServer: <hostname>
>> 2. the Backup server gets the exact same venue configuration as the
>> original, updates are the responsibility of the administrator. No 
>> magic will
>> take care of that for now.
>> 3. The backup server is part of the venuestate that is received by the
>> client upon entering the venue.
>>
>> Behavior:
>>
>> If the venue client detects a break of the event channel, then:
>>
>> 1. it tries 3 times to reconnect to the original server, if that fails
>> 2. it tries to connect to the backup server.
>>
>> Open Issues:
>>
>> How to revert back to the original server?
>>
>> Ideas:
>>
>> 1. don't bother
>> 2. retry at some interval (the trickle of users back would be 
>> disturbing)
>> 3. ???
>>
>> Concerns:
>>
>> 1. Media tools should not stop and restart (it's disruptive)
>>
>> Thoughts?
>>
>> --Ivan
>>
>>
>>
>