Hi, all.

I'm about to start a project to evaluate Ingres Replicator for a
customer who wishes to implement a contingency plan between databases
at two different sites. The idea is relatively simple: from each site,
data will be replicated to an unprotected read-only target database at
the other site.

While I have had substantial experience with "older versions" of
Replicator, it's been a few years since I have really worked with it.
I believe that the last version that I really ever implemented was OI
2.0... which still was based on database rules instead of the newer
change recorder / distribution threads architecture.

I have a couple of questions for anyone with any experience with the
latest versions, and would greatly appreciate any help I can get!

1) The customer's version of Ingres is Ingres II 2.0/0001 (hpb.us5/00).
They are scheduled to migrate to Ingres II 2.5 in November. Are there
any major differences (architectural changes or bug fixes) between
these versions of Replicator? Can we expect any problems with migrating
replicator, or would there be any good reasons to wait and implement
replication only after 2.5 is installed?

2) I know that in the older versions, each "replicator server" must be
compiled and linked within the specific Ingres environment. I assume
this is still true (according to the manual, it still is). My customer
is hesitant to purchase HP-UX's software development kit for what is
initially just an evaluation phase. Does anyone know of another
alternative for C language compilation on an HP-UX 11i box?

3) If anyone has any caveats or recommendations for this kind of
environment, I will gladly receive suggestions! I'm mostly afraid of
possible locking / deadlock problems (which was what I suffered most
with in past versions!). Have there been any improvements in this
area?

Thanks in advance for any assistance!

James Gramling
Commit Work Informatica
Rio de Janeiro, Brasil

On 12 Sep 2005 06:44:32 -0700, jimwgramling (AT) gmail (DOT) com
<jimwgramling (AT) gmail (DOT) com> wrote:
Hi James

The major differences between 2.0 and 2.5 will be that the change
recorder is now built into the DBMS and the replicator server is
generic - and therefore doesn't have to be built.

I confess I've never done a replicator migration. You'd need to drop
all the rules and procedures that are the 2.0 change recorder so I
imagine the easiest way to migrate, especially if your replication
scheme is straightforward, would be to freeze any database changes,
dereplic them, upgrade Ingres and the database and re-configure
replication.

Are you sure you read the 2.5 manual? In any case 2.5 was when we
introduced the generic replication server so you won't need a C
compiler to build one.

See ftp://ftp.ca.com/CAproducts/ingres/d...res_25/rep.pdf

You can now specify the lockmodes for the replication threads dealing
with the distribution and input queues, as well as the shadow and
archive tables. See Chapter 8 of the Replicator manual.

The other thing I would say, is that at this time, I'd be thinking of
upgrading to 2.6 rather than 2.5, if not r3. More features, better
performance and a longer support life.

HTH

--
Paul Mason

Thanks Paul! That's exactly the information that I was looking for.
Eric Mischke had already indicated to me that he thought it was no
longer necessary to have a compiler; but it's great to substantiate it
with a reason why! Is there a "new features" guide or whitepaper for
Replicator 2.5 /2.6 anywhere out there, that specifically targets the
changes from 2.0 to the latest version?

I WAS sure ... but after downloading the 2.5 manual you linked I went
back and double-checked my 2.6 manual in the original 2.6 distribution
tar and -- lo-and-behold -- the Product Requirements specification no
longer mentions ESQL-C and a compiler. My only explanation is that,
when I went to downoad the Ingres II 2.0 rep.pdf, I overwrote the
original rep.pdf file. So when I compared them, they seemed to be
identical (wrt the product requirements).

So this is very good news for the customer; just too bad I didn't know
that and tell them up front! It of course means that it makes no sense
to install replicator now then migrate in November; so we're going to
try and accelerate the upgrade to 2.6 before installing Replicator.

Thanks again for the help!

James Gramling
Commit Work Informática
Rio de Janeiro, Brazil

On 27 Sep 2005 06:26:15 -0700, jimwgramling (AT) gmail (DOT) com
<jimwgramling (AT) gmail (DOT) com> wrote:
Hi,

Sorry for the delay in replying.

I'm not aware of a new features guide for Replicator specifically. I
checked the general one on SupportConnect but it doesn't mention
Replicator and it's fairly brief in any case.

Sorry I can't be of more help.

Regards
--
Paul Mason