ER Replication setup question
Ulf (IDS 11.50.FC7) had two separate Enterprise Replication domains (A-B and C-D) and wanted replication between A and C. Replies confirmed there is no supported cross-domain replication: he must 'cdr delete server' on C and D and re-add them to the A/B domain with -S, then resync. For the firewall concern, he was pointed at the ER topology docs — define A, B, C as root servers and D as a leaf routing all traffic through C — plus the sqlhosts s=6 option to restrict a port to ER/HDR traffic. The thread ends with Ulf's plan and an unanswered question about whether D really needs only C in its sqlhosts, so no final confirmation is recorded.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
Hi ! Version is 11.50.FC7 We are currently using ER replication between servers A and B, and we also have ER setup between servers C and D. I am now trying to add replication also between A and C, but I am not able to find the correct way to do this using cdr define server. A, B are not aware of C and D and vice versa. Any ideas anybody ? Ulf
On 24/09/2010 15:04, Ulf wrote: > Hi ! > > Version is 11.50.FC7 > > We are currently using ER replication between servers A and B, and we > also have ER setup between servers C and D. > > I am now trying to add replication also between A and C, but I am not > able to find the correct way to do this using cdr define server. > > A, B are not aware of C and D and vice versa. > > Any ideas anybody ? > > Ulf No supported way to join "two discrete ER domains" : Domain 1 is A and B Domain 2 is C and D :/ You would have to : "cdr delete server c" on c "cdr delete server d" on d Then you can add c and d into the a and b domain using the -S a to sync to a. Obviously, this will require a bit of thought on "timing" to ensure continued data integrity between c and d. Or perhaps just a "sync"?
Thanks for your answer, I suspected that this was the problem. I do not want to replicate from A and B to D, D is outside a firewall. Is there a way to define that A,B, and C is aware of each other and that D is only aware of C ?
On 24/09/2010 15:58, Ulf wrote: > > Thanks for your answer, I suspected that this was the problem. > > I do not want to replicate from A and B to D, D is outside a firewall. > Is there a way to define that A,B, and C is aware of each other and > that D is only aware of C ? > Yes, you can do this. The following describes the possible topologies : http://publib.boulder.ibm.com/infocenter/idshelp/v115/index.jsp?topic=/com.ibm.erep.doc/ids_erp_100.htm I am not sure what you mean by "aware". I assume that you mean that : A and B and C can all directly see each other, and that only C can communicate with D. You can define ER so that traffic to / from D will all go through C (although I think you do not want to replicate between A and / or B to D, you would probably configure ER up this way).
On 24 Sep, 17:06, "Jonny...@Usenet-News.net" <Jonny...@Usenet- News.net> wrote: > On 24/09/2010 15:58, Ulf wrote: > > > Thanks for your answer, I suspected that this was the problem. > > > I do not want to replicate from A and B to D, D is outside a firewall. > > Is there a way to define that A,B, and C is aware of each other and > > that D is only aware of C ? > > Yes, you can do this. > > The following describes the possible topologies : > > http://publib.boulder.ibm.com/infocenter/idshelp/v115/index.jsp?topic... > > I am not sure what you mean by "aware". > > I assume that you mean that : > > A and B and C can all directly see each other, and that only C can > communicate with D. You can define ER so that traffic to / from D will > all go through C (although I think you do not want to replicate between > A and / or B to D, you would probably configure ER up this way). By "aware" I mean that I do not want to define A,B and C in D's sqlhosts, only C, and also that I only need one hole in the firewall. I am looking at the docs right now and suppose that D would be a "leaf node" and the other "root nodes"
On Sep 24, 9:04 am, Ulf <ulf.akerb...@gmail.com> wrote: > Hi ! > > Version is 11.50.FC7 > > We are currently using ER replication between servers A and B, and we > also have ER setup between servers C and D. > > I am now trying to add replication also between A and C, but I am not > able to find the correct way to do this using cdr define server. > > A, B are not aware of C and D and vice versa. > > Any ideas anybody ? > > Ulf The define server command has to do with defining the domain and routing. It does not have anything to do with what is replicated where. If you want to replicate between A and C, then A and C must be in the same domain. There is no support for cross domain replication.
On Sep 24, 10:06 am, "Jonny...@Usenet-News.net" <Jonny...@Usenet- News.net> wrote: > On 24/09/2010 15:58, Ulf wrote: > > > Thanks for your answer, I suspected that this was the problem. > > > I do not want to replicate from A and B to D, D is outside a firewall. > > Is there a way to define that A,B, and C is aware of each other and > > that D is only aware of C ? > > Yes, you can do this. > > The following describes the possible topologies : > > http://publib.boulder.ibm.com/infocenter/idshelp/v115/index.jsp?topic... > > I am not sure what you mean by "aware". > > I assume that you mean that : > > A and B and C can all directly see each other, and that only C can > communicate with D. You can define ER so that traffic to / from D will > all go through C (although I think you do not want to replicate between > A and / or B to D, you would probably configure ER up this way). You can define a secured port for ER to use by using the s=6 option in the sqlhost file. This will limit the port to ER / HDR traffic only.
Thank you for your help, based on this information this is what I intend to do: Current setup will be deleted Servers A,B and C will be defined as root servers, the will have the possibiltity to connect to each other (they will all be defined in the sqlhosts file for each server and will trust each other using hosts.equiv). Replication will take place between the A -> B and A- >C. I don't see any issues here. Server D will be defined as a leaf server and will not know of any other server than C, meaning that D in its sqlhosts only has information about itself and C. Replication will only take place between C and D. The question I have is if it possible that D only connects to C (D only having C in sqlhosts) ? I do not want to open the firewall more than necessary I have tried to understand the documentation on this issue and it is not clear for me if it works. TIA Ulf