Re: Data lost during replication
From: RB (robert_at_brokerforce.com.NoSpam)
Date: 03/26/04
- Next message: Harmannus: "Replication from code"
- Previous message: RB: "Re: Replica Farm, frequency of interchange?"
- In reply to: Peter Jones: "Data lost during replication"
- Messages sorted by: [ date ] [ thread ]
Date: Fri, 26 Mar 2004 16:26:54 -0700
Peter:
We manage more than 100 replica sets; Loose the dial-up connection to
synchronize. Our experience is that RAS is totally undependable, frequently
crashed the base replica, and was prone to all kinds of synchronizer
problems due to the slow speed. VPN via dial-up is also prone to similar
issues and our databases are generally in the 150 mb range. A dropped
connection on either connection usually damaged the hub.
Use either a LAN connection to direct sync or Internet synchronization.
RB
"Peter Jones" <anonymous@discussions.microsoft.com> wrote in message
news:F3C0A962-AF66-4DD2-BAF1-6190E1A3C409@microsoft.com...
>
> Sorry if this is a double post. I sent the following about six hours ago
and it still hasn't appeared on the forum so I'm sending again...
>
> Hi,
> I'm in pre-deployment testing mode for a replicated system using star
topology over dial-up lines. All my prior testing has been over a LAN and
everything has been fine. My environment is:
> 1) Access 2000
> 2) Windows 2000 IIS Server at the Hub and Windows 2000 on remotes laptops.
Latest SP's installed along with MDAC 2.8.
> 3) The laptops connect directly into LAN via secure dial-up connection,
i.e. not via the Internet.
> 4) The Synchronizer on the Hub is configured for "Indirect
Synchronization". The order of synchronization attempts is: Indirect,
Internet, Direct.
> 5) The Hub database has now been configured to its maximum expected size -
100,000 master records and 1,000,000 child records. It is 300Mb big.
> 6) The large database initially caused problems with running out of locks
and MaxLocksPerFile, in the Jet registry, was increased from 9,000 to
32,000. This change fixed that particular problem.
> 7) A global replica was created and put on a laptop.
> 8) The hub and laptop were synchronized over the LAN using the Access UI
(normally the synchronization is done programmatically).
> 9) 200 master and 800 child records were added at the laptop and the
databases synchronized again over the LAN without problems. The
synchronization was done programmatically and took ~ 1 minute.
> 10) The LAN connection between the two systems was removed and dial-up
connection established (connect speed 21K).
> 11) Again, 200 master and 800 child records added to the laptop database.
> 12) Synchronization was started and left to run overnight (hadn't
completed 1.5 hours after starting).
> 13) This morning the Laptop had stopped with a "delayed write error". This
is a known problem to me that was fixed by turning of write caching on the
disk. However I had forgotten to do that for this test but it is not the
reason for this post.
> 14) On examining the Laptop I can see that, up to point of the "delayed
write error", 27Mb of data had been sent on the dial-up connection and 19Mb
had been received.
> 15) The laptop database was in its original state but, on the Hub
database, about 97,000 master and 970,000 child records had been deleted.
>
> Q1. Why so much network traffic when so few new records had been created
and ready for replication?
>
> Q2. Why were the records deleted from the Hub?
>
> Any comments, suggestions or wild guesses would be appreciated.
>
> Cheeers, Peter
- Next message: Harmannus: "Replication from code"
- Previous message: RB: "Re: Replica Farm, frequency of interchange?"
- In reply to: Peter Jones: "Data lost during replication"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
