Db2 Analytics Accelerator should be catching up changes dynamically on status of HW resources(NVME,OSA,ROCE,Ficon) over LinuxONE when resource come back online after toggle operation.

For LinuxONE systems, MCL or failure of some HW parts sometimes leads us to toggle operation to be able to fix or enable HW resources. What we have noticed is when HW resources (NVME,OSA,ROCE,Ficon) toggled offline over LinuxONE, Db2 Analytics Accelerator marks it as red over Admin UI. But sadly after change is performed if we want to toggle that resource back online Admin UI doesn't show it as available and requires deactivation/activation of all Accelerators which are using that affected resource. Unfortunately that means each time service needs to go down and start back again. When considering for production environment with lots of resources(CPU,memory,..) that task takes some time for system to finish. Addition to that requires someone to perform deact/act over HMC

What we used to have on z systems is, this kind of small toggle actions can be handled dynamically without any affect or any additional actions to take. With some zOS commands we have a possibility to manage change and eliminate any potential affects of toggle operation. I am aware SSC is different running environment than z/OS lpars but somehow if we have this feature on new versions that would relieve us from burden of deactivation/activation of each accelerator even for the small operation.

Needed By

Not sure -- Just thought it was cool

Post comment

Admin
Mohamed Elmougi

Jun 1, 2026
Hi Yasin, thanks for opening the IDEA and here is our summary of our reviewing process:
- OSA and ROCE: If a network device goes down or replaced, depending on what network device this is, this might create a temporary outage. IDAA team has validated that with IDAA Version 8, the system comes up working again after the network is restored. On the level we are operating, we do not need to look into the differences between OSA, RoCE, Network Express and other variations of network
- FICON storage: we typically use this for persistent storage, and we do not tolerate any inconsistency with the storage layer. If part of the storage goes down, we need a proper restart. With our architecture and design, it is close to impossible to let us survive temporary storage failures, at least for the persistent storage. We expect to run in enterprise storage environments, with all the usual extra redundancy for storage from the hardware side. If for some reason part of our storage pools becomes unavailable, we consider this fatal. To protect from such an event, we encourage considering HA setup or GDPS integration. As usual, this can also help for scheduled maintenance of the storage environment.
- For transient storage (including NVMe), it might be possible for us to extend the product in a way we survive a temporary outage of some transient storage devices, that would be a line item with significant effort.
As the idea considered several items with different status. Currently we will put the item to not under consideration. And we encourage you to open another IDEA for Transient Storage (NVMe) in which we will need more clarity and information on what can be satisfying the expectations.
Reply
Hide replies

Guest

Feb 3, 2026

Small update; if i need to prioritize HW resources for this functionality considering frequency of failure, toggle operation requirement after MCL, number of parts we are using and importance that would be:
1- FICON and OSA
2- ROCE
3- NVME
FICON and OSA are the most needed for this functionality
Regards
Yasin

Reply
Hide replies

By clicking the "Post Comment" or "Submit Idea" button, you are agreeing to the IBM Ideas Portal Terms of Use.
Do not place IBM confidential, company confidential, or personal information into any field.

Please enter your email address

RELATED IDEAS

Db2 Analytics Accelerator should be catching up changes dynamically on status of HW resources(NVME,OSA,ROCE,Ficon) over LinuxONE when resource come back online after toggle operation.