A.15 Limitations of Automated Server Group and Automated Site Upgrade

For multi-active server groups, such as DA-MP/vSTP MPs, non-deterministic server selection could possibly result in a network outage during the upgrade. In certain scenarios, customer preferences or requirements can result in configurations in which it is imperative that DA-MP servers must be, or conversely, cannot be, upgraded together. These scenarios are described in this section with the recommendation that customers do not use ASG if any of these exists in their network.

Note:

Oracle’s recommendation for any customer whose network aligns with any of the following scenarios is that the Automated Server Group upgrade should NOT be used on multi-active DA-MP server groups. Use of ASG risks a potential network outage. For Automated Site Upgrade, following limitations can be solved by rearranging/adding the upgrade cycles. If the user does not want to create a custom upgrade plan by rearranging/adding cycles then in that case manual upgrade procedure from section 5.3 method should be used.

Specialized Fixed Diameter Connections

In this scenario, each peer node is configured to connect to two specific DA-MPs for local redundancy (Figure 18). With ASG/ASU setup for 50% minimum availability, three of the DA-MPs in the server group are upgraded in parallel. However, it is not possible to determine in advance which three DA-MPs are selected. Although the DSR has redundant connections to the peer nodes, an unfortunate selection of servers for upgrade could result in an outage. Upgrade cycle 1 takes out both DA-MPs connected to the unhappy peer. This peer is isolated for the duration of the upgrade. The happy peer is connected to DA-MPs that are selected by ASG/ASU for different upgrade cycles. This peer is never isolated during the upgrade.

Specialized Floating Diameter Connections

In this scenario, each peer node is configured to connect to an IPFE TSA address hosted by a set of DA MPs. When any particular TSA contains only a subset of the server group MPs, and the DSR upgrade logic happens to select that subset of MPs for simultaneous upgrade, then there is a signaling outage for that TSA. This scenario is depicted in Figure 19. TSA1 is distributed across the first three DA-MPs, whereas TSA2 is distributed across all six DA-MPs. If ASG/ASU is initiated with 50% minimum availability, the DSR could select all three of the DA-MPs hosting TSA1 in the first upgrade cycle. The unhappy peer is isolated for the duration of upgrade cycle 1. The happy peer is connected to TSA2, which is hosted by the DA-MP servers in such a way that the TSA is evenly hosted in both upgrade cycles. This peer is never isolated during the upgrade.

Specialized Distribution of DSR Features

In this scenario, the customer has decided to enable P-DRA and RBAR on four DA-MP servers and DCA on two DA-MP servers, consistent with expected traffic load. With ASG setup for 50% minimum availability, the DA-MP server group is upgraded in two cycles. RBAR and P-DRA happen to be hosted by DA-MP servers selected by ASG/ASU to be in different upgrade cycles, albeit unbalanced. The RBAR peer is only marginally happy because during upgrade cycle 1, only 25% of RBAR and P-DRA capacity is available, even though the customer specified 50% availability. DCA happens to be hosted by DA-MP servers selected by ASG/ASU to be in upgrade cycle 2. In this event, the DCA peer is unhappy because DCA is completely unavailable during upgrade cycle 2.