IBM Domino clustering was introduced approximately 20 years ago to provide IBM Domino applications with a transparent high availability and scalability solution. It has been a core and critical part of essentially all production IBM Domino deployments since that time.
Over the years, numerous improvements have been made in IBM Domino clustering - Streaming Cluster Replication being a particularly notable example. But, with IBM Domino 10 we will be adding the most significant set of improvements to IBM Domino clustering since its original introduction.
At a high level IBM Domino clusters are trying to provide the illusion of a single resource to applications - whether those are being accessed from IBM Notes, iNotes, Verse or any other application. In this post I'll be focusing on how we are improving the content handling aspect clusters for administrators and end users.
For an administrator trying to insure content equality across a set of IBM Domino servers in a cluster there are a number of challenges including:
But even if administrator practices are exceptional, end users still sometimes can see through. Delayed replication between IBM Domino servers in a cluster can result in different users seeing different content briefly. In some cases a single user may see content "disappear" as the result of delayed replication. And, having IBM Domino be restarted for any reason in a clustered server can exacerbate these situations.
So, for IBM Domino 10 we are offering a set of new features to help both administrators and end users in all of these dimensions. Some of the key new features are:
Automatic Symmetric IBM Domino Clusters
Administrators will be able to declare one or more folders (a.k.a. directories) of databases should be automatically kept consistent across servers in a cluster. For example, IBM Domino servers in the Sales cluster may have a folder named "Leads" which contains a set of databases. Once done, each IBM Domino server will be periodically look to see whether it is missing content present in any of those specific folders on any other cluster in the server and will automatically create the needed databases for itself. This automatically handles a number of situations quickly and automatically:
Removing IBM Domino Server Restart Problems
In steady state when all servers in a cluster are up and running, IBM Domino's Streaming Cluster Replicator will insure that content is shared across cluster members extremely quickly. But, before IBM Domino 10, when a IBM Domino server was restarted, it would take some time for the Streaming Cluster Replicators in the cluster to get back to this steady state. How long depended on how many databases had to be checked and how much changed content had accumulated on the other servers in the cluster before the restart completed and replication could resume.
New with IBM Domino 10, this restart phenomena will be smoothed out and made essentially transparent.
First, IBM Domino will periodically snapshot knowledge of Streaming Cluster Replicator state on each server. That will allow the Streaming Cluster Replicator on the restarting server to dramatically reduce and often completely eliminate any work needed to get back to steady state. Because SCR is typically sending updates to the other server in its cluster so rapidly, there is often no work for it to do once it restarts. Recording this information on disk helps us take advantage of that directly.
Second, IBM Domino servers other than the one restarting will wait for a period of time, continuing to queue updates to be sent to the restarting server rather than reacting abruptly. In most cases, a routine IBM Domino restart or OS reboot of a server will take less than the wait period - resulting in this accumulated content quickly getting pushed to the restarted server and made available to users.
Third, the IBM Domino server that is restarting may delay accepting client connections for a short period of time to allow content to be brought up to date to avoid end users from seeing or even acting on out-of-date content. Enablement of this feature and the maximum period of time the server will delay accepting connections will be administrator controllable.
Replication Currency Monitoring
With IBM Domino 10, administrators will be given monitoring statistics to allow direct tracking of latency - allowing them to have a clear picture, to be alerted and to be able to respond in real time to any content currency issues which may impact their end users. Combined with IBM Domino 10's ability to directly publish statistics to internal or external services, administrators can have a new level of alerting and visualization of data currency on their servers.
We believe these features will make a significant improvement to administrator and user experiences but we have plans for many more improvements - most of which came to us via the IBM Domino 2025 Jam input. Stay tuned for further updates.