Thanks for the great info Martin.

If you don't mind me asking, what technology do you use for your cache?  How do you combine the data in a local site's cache with the (potentially different) data in that local site's slave DB?

Thanks,

--S

-----Original Message-----
From: slony1-general-bounces at lists.slony.info [mailto:slony1-general-bounces at lists.slony.info] On Behalf Of Martin Eriksson
Sent: Friday, August 22, 2008 12:54 AM
Cc: slony1-general at lists.slony.info
Subject: Re: [Slony1-general] Slony over a WAN?

Hi, we do slony over WAN where W = World hehe

Our master Database is in London, then we got one node in Cyprus, one 
node in San Fransisco, and then one node in Frankfurt and about to add 
another node in Hong Kong. Our database is only around 8 gigs though.

we got a bit of a special setup in terms of the applications using the 
dbs as we want our application to read locally but write to the master 
so we a pretty advanced cache system for handling if Bill is sitting In 
San Fransisco writes to the db and then look at the data he sees the 
data he just wrote so does everyone else in that office even though it 
might not have been replicated to his local db yet. this work pretty 
good, we never have more then 5-15 seconds delay on slony event across 
the globe so in worst case if someone tries to change something that 
already been changed it wont work and they just reset their cache (which 
takes 2 minutes) and then they continue working.

and we do have an AWFUL line london <-> Cyprus that on average does 
40kbytes/s which is horribly low. But it still works 100 times better 
then having both read and write going to London all the time from across 
the world..

Of course if the shit hits the fan so to say when we do DLL changes 
which happens every 3 months for release and we cant recover it will 
take up to 72 hours to replicate to all nodes (which is not really an 
option)

but if you got good bandwidth between your nodes its not a problem.

Though if you got a 80 gig db you might want to consider not replicating 
it across the WAN as it will take quite a long time and might cost your 
a lot depending on your ISP i guess.. well if you got a gigabit line and 
dont mind using it then I guess you are fine :) but if you bandwidth is 
limited you could do as we do sometime,

setup a second db instance on your masternode (assuming node and master 
will be in the same hardware architecture) do a replication to the local 
instance once done shut down that instance move the whole /data 
directory of that instance onto so movable disk, and drive down to the 
other node load it up, modify path to the node using slony store path 
and fire it up and let it catch up on the last hour or so of the new data.

 slon processes should run in the same "network context" as the node that
each is responsible for managing so that the connection to that node is a
"local" one. 
     Do not run such links across a WAN.



Its already been covered but I'll add to it. Yes you should run the slon 
process on the node in question do not run them all on the master node, 
not only because it will work better also because when you do have 
multiple slon process running on one machine it can get VERY confusing 
figure out which one goes where and which one is having connections 
open. Life is much easier if one slon process run on each node machine.

good luck!
Martin


Mark Steben wrote:
> Looking for advice on how to proceed.  We are running Postgres 8.2.5 in Lee,
> Massachusetts.  We will be installing same in Norfolk Virginia in the not
> too distant future.  Our need is to replicate roughly 60 - 70 percent
> Of our 80 GB database in Lee over a WAN to Norfolk for reporting purposes.
> In reading 'Slony-1 Best Practices' on the Website I came across the
> following statement:
>      
>   slon processes should run in the same "network context" as the node that
> each is responsible for managing so that the connection to that node is a
> "local" one. 
>      Do not run such links across a WAN.
>
> Does this still hold true?  If not I would like to hear experiences of
> anyone engaging in a 'Slony-1 long distance relationship'
> Any other alternatives to consider?  I do run Slony-1.2.14 in development
> with everything encompassed in Lee.  However 
> we will be opening another office 35 miles west in West Springfield that I
> will be operating out of.  I plan on employing Slony-1
> To provide replication between these 2 'not so long distance' locations.
>
> Any comments welcome.  Thanks
>
>
> Mark Steben│Database Administrator│ @utoRevenueT
> 480 Pleasant Street, Suite B200, Lee, MA 01238 
> 413-243-4800 x1512 (Phone) │ 413-243-4809 (Fax)
> A Division of Dominion Enterprises
>
>  
>  
>  
>  
>  
>
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at lists.slony.info
> http://lists.slony.info/mailman/listinfo/slony1-general
>   

_______________________________________________
Slony1-general mailing list
Slony1-general at lists.slony.info
http://lists.slony.info/mailman/listinfo/slony1-general