Question Re: CI and ASCS instance in Technology Q&A

CI and ASCS instance

Former Member — Sat, 05 Oct 2013 09:25:49 GMT

In my HA installation , If I have CI which is going to failover to backup node , do I still need to instance ASCS instance ?

Re: CI and ASCS instance

Former Member — Sat, 05 Oct 2013 12:38:56 GMT

Hello,

CI is no more single point of failure. It is ASCS instance. You need to install ASCS install so that it can move across 2 nodes in HA installation.

CI will be a local installation and there should not be failover for CI.

Thanks,

Sunny

Re: CI and ASCS instance

S_Sriram — Sat, 05 Oct 2013 13:43:23 GMT

Hi Rao

Could you share some error log? Some of the HA SCN refer links for troubleshooting propose

http://scn.sap.com/people/benny.schaich-lebek/blog/2009/11/23/high-availability-with-sap

http://scn.sap.com/docs/DOC-8541

http://scn.sap.com/docs/DOC-7848

Thanks

Sriram

Re: CI and ASCS instance

Valdecir — Wed, 19 Mar 2014 18:53:36 GMT

Hi Rao,

In older times SAP used to have only CI and it was commonly configured to have HA. With this kind of installation you always have end users connected to the CI and when you need to bounce the message server and/or enqueue server, all users connected to the CI would be kicked-off the system, causing an impact to those end users.

To improve this scenario, SAP started to deliver the new installations using the ASCS concept which is a split of the former CI (also known as DVEBMGSxx) into two new instances: ASCS and Dxx.

ASCS now stands for message/enqueue servers only (no end user connects to it) while Dxx is responsible for all other WP types ( DIA, BTC, UPD, etc). Dxx is called Primary Application Server or PAS and additional servers are called Secondary Application Servers or SAS.

With this approach, you can bounce ASCS without affecting the end users. If you have a PAS and at least one SAS then you can create a logon group with them. In case you need to bounce one of them, you just need to remove that particular server from the logon group, wait for all users to disconnect (which will happen naturally) and then bounce the server. Once it is online again, just put it back to the logon group.

Considering the HA scenario, it is MUCH faster for tha HA structure to re-start only the ASCS which has only two services.

So, if you want to implement HA you can do it with your current implementation, however we recommend you to split it to have the ASCS and have your HA much faster.

regards,

Valdecir