Hello.
I searched old posts for a possible solution to the problem I have but
without success.
The summary is this: is there a possibility to continously restore
transactions commited on one Informix instance to another?
I know about HDR and Enterprise Replication but each one is inadequate, each
for its own reasons.
HDR would require the secondary server to be online all the time (which may
not be true in our case), and Enterprise Replication isn't too much reliable
at transaction level, is it?
The configuration that we use now is the following: we have three Sun
machines, each running one Informix instance (9.40). One is production
machine, and the other two serve for testing and development purposes, and
we would also like to use them in a failover solution. Production instance
is backed up using onbar and Legato Networker software to a backup server
containing bunch of tapes. Dbobjects are backed up every night, and logical
logs are backed up continously using log_full.sh script.
What we'd like to achieve is this: every time a log on a production instance
is filled and backed up on a backup server - replay that logical log on
other two instances. That would require to trigger a process on those
failover instances every time a logical log fills on a production instance.
Then that process would have to connect to backup server and apply latest
logical log.

I'm not a Informix guru so maybe these ideas may sound ignorant.
I'd like to hear if anyone implemented something like this.
Of course, I'd appreciate to hear any other solution with the same or
similar effects.
Thanks!

Davorin Kremenjas wrote:

> Hello.
> I searched old posts for a possible solution to the problem I have but
> without success.
> The summary is this: is there a possibility to continously restore
> transactions commited on one Informix instance to another?
> I know about HDR and Enterprise Replication but each one is inadequate, each
> for its own reasons.
> HDR would require the secondary server to be online all the time (which may
> not be true in our case), and Enterprise Replication isn't too much reliable
> at transaction level, is it?
> The configuration that we use now is the following: we have three Sun
> machines, each running one Informix instance (9.40). One is production
> machine, and the other two serve for testing and development purposes, and
> we would also like to use them in a failover solution. Production instance
> is backed up using onbar and Legato Networker software to a backup server
> containing bunch of tapes. Dbobjects are backed up every night, and logical
> logs are backed up continously using log_full.sh script.
> What we'd like to achieve is this: every time a log on a production instance
> is filled and backed up on a backup server - replay that logical log on
> other two instances. That would require to trigger a process on those
> failover instances every time a logical log fills on a production instance.
> Then that process would have to connect to backup server and apply latest
> logical log.
>
> I'm not a Informix guru so maybe these ideas may sound ignorant.
> I'd like to hear if anyone implemented something like this.
> Of course, I'd appreciate to hear any other solution with the same or
> similar effects.
> Thanks!
>
>

It would be possible if you'd maintain the instances in rollforward...
This means they are not accessible by clients.

I suppose this is not what you want to do...

"Davorin Kremenjas" <davorin.kremenjas@inet.hr> wrote in message
news:c6qvtv$ffe$1@bagan.srce.hr...
> Hello.
> I searched old posts for a possible solution to the problem I have but
> without success.
> The summary is this: is there a possibility to continously restore
> transactions commited on one Informix instance to another?
> I know about HDR and Enterprise Replication but each one is inadequate,
each
> for its own reasons.
> HDR would require the secondary server to be online all the time (which
may
> not be true in our case), and Enterprise Replication isn't too much
reliable
> at transaction level, is it?

yes, that's what it's built for...

> It would be possible if you'd maintain the instances in rollforward...
> This means they are not accessible by clients.
> I suppose this is not what you want to do...

Well, they don't necessarily have to be accessible by clients.
That solution would also suffice.
Can you give me some more info please, URL or something?