|
|
Our Hot Pick: Rising Antivirus 2006 - Certified by TUV & Checkmark! Get 10% discount by entering this coupon code: ONDISCOUNT10
I have read the Chapman/Kaplan article, the MSDN article Database
Replication in Microsoft Jet 4.0 and various postings about replication.
Still a few [more] questions.
I have the TSI synch object installed and working with the MDB applet
that I downloaded from Kaplan's site. I also have replman installed and
working.
My planned implementation
-------------------------
My target environment will be about 25 laptop users and about 20 LAN
connected desktops split into two locations connected by a partial T1.
I plan to have a replica at each location that each laptop replicates
with when they are in the office and connect to the LAN. I also need a
replica at each location that the LAN attached desktops connect to.
Q1 - Can these be the same replica?
I also need a replica farm at each end, probably three replicas in each
farm, that are locally updated regularly. Also scheduled updates
between the two farms via indirect replication over the WAN.
Q2 - Can I use one of the replica farm members for
a)access by LAN attached PCs?
b)the replica that laptops sync to when connected to the LAN?
It has also been recommended that on each laptop I have a local backup
in terms of a replica that is synchronized each time the application on
the laptop is closed.
Q3 - Anything about the above approach you would suggest I change?
Other questions
-------------------
Q4) In ReplMan, when I view all replicas, it shows two unmanaged partial
replicas. I don't recall creating ANY partial replicas.
Q5) I still struggle with the definitions of and differences between
managed/unmanaged and scheduled/unscheduled. Can someone define them
well for me?
Q6) I can use the Access applet that interfaces with the TSI
Synchronizer object. I also have Replman working. And yes I have a
synchronizer that runs. I am still unclear on how you define and name
synchronizers. I think I recall naming the synchronizer or accepting
the suggested name. I have one on this PC. How should I properly
create and name a synchronizer?
Q7)When do I need a 2nd synchronizer on a single PC and how do I create it?
Thanks again.
Bob
|
|
user[ at ]domain.invalid wrote in
news:uodDdJ6yGHA.4368[ at ]TK2MSFTNGP02.phx.gbl:
[Quoted Text] > I have read the Chapman/Kaplan article, the MSDN article Database
> Replication in Microsoft Jet 4.0 and various postings about
> replication.
> Still a few [more] questions.
>
> I have the TSI synch object installed and working with the MDB
> applet that I downloaded from Kaplan's site. I also have replman
> installed and working.
>
> My planned implementation
> -------------------------
> My target environment will be about 25 laptop users and about 20
> LAN connected desktops split into two locations connected by a
> partial T1. I plan to have a replica at each location that each
> laptop replicates with when they are in the office and connect to
> the LAN. I also need a replica at each location that the LAN
> attached desktops connect to. Q1 - Can these be the same replica?
- Are the laptops using direct or indirect synch when connected to
the LAN (I'd use direct)?
- how often do these synchs happen?
- do you have bound memo fields in your application?
A synch that takes place while a user is editing can cause these two
problems, the first transient, the second fatal to the record
involved:
1. a design change on a table that is open (whether being edited of
not) will fail. This problem will vanish as soon as a synch happens
when no one has the table open.
2. a data edit propagated to a record that has a memo field that is
being edited at the time the synch takes place will corrupt the memo
pointer and cause the loss of the data record (though everything but
the memo can be manually recovered). This can be avoided in 99.9% of
cases by using unbound fields for editing memos. I've explained how
this is done in this newsgroup many times (look for "unbound memo"
under my email address in Google Groups for this newsgroup).
I always use one replica for synching and one for editing.
However, I never had arbitrary synch times, since I was only ever
synching between two locations with LAN users editing a single
replica. That meant that the time to schedule the synch between the
hub replica used for the synch and the production replica used for
edits by the LAN users was knowable. If you have laptop users
synching with the hub replica at arbitrary times, you'll probably
want to use indirect replication to synch between the production
replica and hub replica before they synch direct with the hub, and
then an indirect synch back from the hub to the production replica.
> I also need a replica farm at each end, probably three replicas in
> each farm, that are locally updated regularly. Also scheduled
> updates between the two farms via indirect replication over the
> WAN.
>
> Q2 - Can I use one of the replica farm members for
> a)access by LAN attached PCs?
The purpose of a replica farm as conceived by Michael Kaplan (it was
his idea and his term) was to make a synch with a synchronizer
fail-proof. When you initiate an indirect synch from a remote
location, you synch with a synchronizer, which chooses among its
managed replicas which actual replica is the partner in the synch
with the remote replica. By having multiple replicas managed by the
synchronizer, if something happens to the replica that was last
used, it just picks the next available replica.
I would never manage my production replica, because that would mean
that you couldn't forecast when it would be synched with. But I'm
not sure there's any real reason why that's an issue.
> b)the replica that laptops sync to when connected to the LAN?
Not sure on this one. I'd be likely to make it a separate replca,
because that increases redundancy. A managed replica on a synch is
going to be more involved, and I'd use direct for the LAN synchs
from the laptops. If you're using indirect, then the question
vanishes, as you'd be synching with the synchronizer, rather than
choosing which replica to synch with.
> It has also been recommended that on each laptop I have a local
> backup in terms of a replica that is synchronized each time the
> application on the laptop is closed.
That's me that recommends that. The idea is that if the edited
replica gets corrupted or lost, you've got a good chance of losing
very little data.
> Q3 - Anything about the above approach you would suggest I change?
>
> Other questions
> -------------------
> Q4) In ReplMan, when I view all replicas, it shows two unmanaged
> partial replicas. I don't recall creating ANY partial replicas.
I don't know, but you can try synching with them to delete them from
the replica set.
> Q5) I still struggle with the definitions of and differences
> between managed/unmanaged and scheduled/unscheduled. Can someone
> define them well for me?
The only replicas that you want managed are the ones that you want
to be controlled by the synchronizer. On a practical basis, that
means that you manage only the replicas that you want to potentially
be a partner in a synch with a remote replica. Keep in mind that a
managed replica is always open, so it would be impossible to compact
it while the synchronizer is running.
The upshot of this is that I would never manage:
1. production replica that LAN users edit.
2. hub replica that is used in a direct synch with LAN laptop users.
Schedules can apply to both managed and unmanaged replicas, as long
as one of the partners in the scheduled synch is managed.
To me, your scenario should be a dual star topology, with two
synchronizers at the center of the two stars, and the hub and
production replicas as the points of the two stars. In a direct LAN
synch topology, the hub replica then becames the center of a
subsidiary star with the laptops as the points of the star. If,
however, you want to do an indirect synch from the laptops to get
the benefit of the replica farm, the laptops would be points on the
star centered on the synchronizer.
> Q6) I can use the Access applet that interfaces with the TSI
> Synchronizer object. I also have Replman working. And yes I have
> a synchronizer that runs. I am still unclear on how you define
> and name synchronizers. . . .
There is one synchronizer per machine, since only one copy of the
synchronizer can be running at once, and the usual is to accept the
name of the machine as the name of the synchronizer. I don't know
that this can be changed.
> I think I recall naming the synchronizer or accepting
> the suggested name. I have one on this PC. How should I properly
> create and name a synchronizer?
I don't believe that you have any control over this. Once you start
up the synchronizer, it exists and has the name of the machine it's
running on. As far as which synchronizer is managing a replica, that
is determined by which synchronizer you use to manage the replica.
> Q7)When do I need a 2nd synchronizer on a single PC and how do I
> create it?
That is impossible, and makes no sense. A single synchronizer can
manage as many replica sets as you need, though only if both replica
sets use the same workgroup file (or a workgroup file that is a
superset of the workgroups for the different replica sets).
--
David W. Fenton http://www.dfenton.com/
usenet at dfenton dot com http://www.dfenton.com/DFA/
|
|
|