on 11-11-2014 1:56 PM
Dear friends, we are planning an HA instalation, but we have a doubt about the Enqueue Replication Server. As per my understanding, we can install en enqueue replication server (ERS) in other host (B), but in case of fail over, the sap instance needs to be started in the same host of the ERS (B) to be able to access to the lock table of the ERS.
Do you know if posible to instal the ERS in other host, and the sap instance in another? The scenario looks like:
Before Fail Over:
Host A: ASCS + CI.
Host B: Stand by Server for the failover.
Host C: ERS.
After Fail over:
Host A: Stand by Server for the failover.
Host B: ASCS + CI.
Host C: ERS.
Will be possible for the host B read the replicated lock table in the host C after the fail over?
The documentation says that the replicated lock table is stored in a shared memory area, then, when the instance is started in the other host (failover) is cappable to read this information form the shared memory area. But what happens if the ERS is in a separated host?
Thanks and best regards.
Hector.
Hi Hector,
The Basic of High availability is to distribute SAP Single point of Failures to nodes which can fail over.
For SAP, the SPOFs are Enqueue and Database. in High availability installation, the server is divided into ASCS which has Enqueue if installing on Unix(ERS is separate in Windows cluster configuration), DB, CI and Dialog instance.
so ASCS and DB should be on separate nodes which can fail over.
With the config you gave above, what if your host C goes down, your system will collapse.
Its better to keep ERS in ASCS and distribute DB and ASCS on separate nodes.
AC
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
I agree with Alberto here in that the enqueue process needs access to the replicated lock informatin in order to restart cleanly and keep the SAP application locks valid. So yes the ASCS should start on a node where the ERS runs.
RE: Amit's comment perhaps I don't understand it but the ASCS and ERS should be split. If the ERS is maintaining replicated lock information and runs on the same node as the ASCS, and therefore the enqueue process, then if the node goes down both the primary and replicated lock information will be lost which defeats the point of having an ERS in the first place.
There is an option to store lock information at the file system level but this is not recommended due to latency on I/O. Never seen this be used.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello Hector,
the ASCS needs to attach to the enqueue replication table (in shared memory) after the fail over in order to keep the locks.
For this reason, having an ERS in a host where ASCS will never run has no sense.
Thanks and KR,
Alberto
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Your ASCS instance must follow ERS instance during failover. In any other cases you lost enqueue table.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Hector,
I have also used similar scenario as mentioned by Tomislav,
PAS (Primary Appl Server) instance number to 10
ASCS instance number to 11
Install ERS for the SCS instance to 12
I have used two Hosts A and B with All ASCS, PAS, SCS, ERS for ABAP+JAVA
Thanks!!
regards,
Abhishek
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello Hector,
I install HA environment with two nodes with ASCS, ERS, D01 and D02.
All 4 instances can switch between those two nodes and in all scenarios system working properly.
You just need to install all instances on VIP which also must be able to switch between nodes together with instances.
Br
Tomislav
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Tomislav,
Thank you for your reply.
Then you said, "to use separate VIP for each instance.". what is each instance? I mean the instance is ERs and ASCS? or the each instance indicate to ASCS instances?
I think you indicated the instances are ERs and ASCS. then Why does ERs need a virtual IP address? because as I know ERs connect to ASCS. so I think ERs doesn't need a virtual IP. but if there has ERs client, the ERs must have a virtual IP for connection. is it right?
Regards,
Jinkoo
Hello,
The ERS must be running on a different host.
Otherwise, having an ERS is not helpful.
Read High Availability with the Standalone Enqueue Server - Standalone Enqueue Server - SAP Library .
Cheers!
Isaías
User | Count |
---|---|
88 | |
10 | |
10 | |
9 | |
7 | |
7 | |
6 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.