I’ve asked this question in the ADABAS Cluster forum, but got no response…
Is it possible to run multiple instances of the same ADABAS database on different cluster nodes sharing that same physical database?
Obvioulsy, I know its not currently possible. I guess I’m asking what are Software AG long term plans to support ADABAS (and NET-WORK) running in a high-availability cluster?
Running ADABAS in a high-availability cluster is possible; there are customers who are doing this. However, there is no direct support for this in the product, as it is available, for example, for Tamino - the customer himself is responsible to create the necessary scripts for a correct failover processing.
A load balancing as on mainframe is not possible in a cluster. You can only start different databases on different cluster nodes.
Yes, I am aware that ADABAS can run in a high-availability cluster. We are doing that now with HP ServiceGuard cluster, but strickly in a “failover” scenerio. I can not start the same ADABAS database on each of the nodes in the cluster all sharing a single physical copy of the database. I run each database only once and if it fails I can restart it on another node.
So I ask again, are there any plans to run ADABAS as a true cluster database?
It is not planned for Adabas v6.1. The features to be implemented afterwards are open - it depends on how many customers request a feature. If cluster database on Open Systems is an important feature for you, please contact the Software AG marketing. If there are enough other customers requesting this feature it may be implemented in a following version.
It’s been a year… has there been any update on ADABAS Cluster Services for Open Systems?
We are currently planning the functionality for v6.2; at the moment the status for Adabas cluster services with load balancing is “after v6.2”; only failover clusters are supported.
We have our own failover cababilties with ADABAS 5.1.4 using Hewlett-Packard’s Service Guard and scripts I have written. Where I have difficulty is when the failure occurs with ADABAS because I need to understand why ADABAS failed and act accordingly. In most cases, auto-restart takes care of the situation without failing over to another server. When auto-restart fails, switching servers doesn’t help because its still the same physical database.
So I’m very interested to know what capabilities are you implementing in ADABAS 6.2 that can assist with and ADABAS failure situation?
I’m also very interested in true Cluster Services with load balancing between multiple database nodes to improve performance and as I see it help guard against physical database node failures.
Also, has there been any progress to allow me to perform database maintenance in parallel with online users? For example, I need to add new database field or invert a new descriptor without locking the file which issues a BT and response code 9 or a response code 17 to the user.
All your requirements are useful. However, we have a long list of requirements, and we must prioritize them, but this is task of our marketing. Therefore I can only recommend, as I already did above, to contact our marketing.
Our high availability solution is based on Veritas, while you are using Hewlett-Packard’s Service Guard. Therefore it may be not appropriate for you. Nevertheless the basic problems should be the same; perhaps it is possible to arrange a meeting with one of our HA specialists. You should contact our marketing for this, too.
You ask, if there has been any progress in utility usage in parallel to online users. Since v5.1 it is possible to perform read-operations in parallel to ADAINV INVERT.
I have talked to marketing about my requirements for ADABAS Cluster Services on Open Systems. They are trying to figure out what to do, what can be done, if R&D is even willing or able to develop Cluster Services for Open Systems, and if there are alternative solutions. Everything is on the table so to speak. Even Oracle has been mentioned. We are desparate for a solution. I just don’t know how much longer we can wait on Software AG to do something.
As for running ADAINV in parallel, I don’t recall reading that bit of news in the release notes or manual. But none the less, I don’t have many files that are read-only access. And as long as a single user is updating the file, I am unable to invert a new descriptor; %ADAINV-E-NOTAVL, File 50 not available. If I lock the file, then all update users, including users just reading the file, are “locked out” and their transactions backed out. It’s very ugly and is hardly the “in parellel” solution I require. An update file lock that only affects users that are actually updating the file could be helpful… Anyway, I’ve been talking to marketing for the last few years about the need for utilities to run “in parallel” with online users as well. It hasn’t helped.
I think we should close this discussion now. I can understand very well your frustration that you request improvements in Adabas, and nothing happens, but Adabas development does not decide on the priorities of new features to be developed. I can only verify that your points are really included in the list of open requirements for Adabas Open Systems.
I understand. The direction of this conversation has gone beyond the scope of this public forum. No one wants their dirty laundry aired in public.
I am happy to continue this conversation in private if you wish, if not I will check back next year to see if any progress has been made.