Power Global

Power Global

A central meeting place for IBM Power. Connect, ask questions, share ideas, and explore the full spectrum of Power technologies across workloads, industries, and use cases.


#TechXchangePresenter
#Power

 View Only

Power11 Lessons Learned

By Shawn Bodily posted Wed April 01, 2026 10:17 PM

  

Below are some recent experiences/issues I've had installing new Power11 systems.




1124s (dont expect it matters if S1124 or L1124)
Post install of L1124s customer had ethernet adapter ports constantly flapping up/down on each/all of these model system in their environment. They do have the same card types/SFPs in their 1122s with no such problem. They checked the switch, replaced the SFP on the switch, replaced them on the server and moved the card and the problem only presented itself on the card in a specific card slot (C8).   They also observed that the SFP was very hot to the touch on that card when replacing them. They decided to move the cable from C8-T0 to C9-T0.   Flapping stopped.  After a service call/ticket with was opened IBM reports it's an airflow situation with the fan and is working on a firmware upgrade scheduled out in June.



E1150 
While performing an install 2 of the 3 E1150s came with mismatched rails. It appears (to me based on E1050 experiences) the wider one is from an E1050 and the thinner one is new to the E1150. One E1150 did come with both thinner rails.  Now, they did work and seem to be compatible but just odd and suspect its simply a parts shortage.





image




Unsure if it's related to the combo of the rails, but when sliding it into the rack the bottom back right corner about 1/32" was rubbing the vertical rail in the rack. We had to pop a handle back on near it it give it a nudge. Speaking of the handles, though a set ships with every system, and they are disposable as they are only attached to load the system, its a good idea to keep one complete set handy. So years down the road when they are removed from the rack it can be done so in a similar fashion. Just reuse the one set as needed.



HMC v11 slowness
A new CR2 with 64GB memory and it manages ONLY these 3 E1150s, no configs/LPARs but seems sluggish in responsiveness. Everything goes dark gray lines...wait wait wait...then refresh.. then it populates. Tasks hang. Initial response from support was v11 code should run on HMCs with 128GB of memory. However upon further review turns out it’s a known problem in V11 R1 SP1111 and fix MG71757 indeed addresses it. 



https://public.dhe.ibm.com/software/server/hmc/fixes/MF71753_ppc.iso

6 comments
61 views

Permalink

https://community.ibm.com/community/user/blogs/shawn-bodily/2026/04/01/power11-lessons-learned

Comments

Carlo Castillo
Mon April 06, 2026 05:51 AM

You know what? I observed the same thing with a new CR2 with 64GB.  But the GUI has no problems at all when you access the HMC remotely.  It's only on the HMC itself that the GUI runs (walks?) slowly.

I should check that 717757 fix.  Thanks for the tip.

Shawn Bodily
Thu April 02, 2026 11:43 AM

Dual and new HMCs, both exhibit same behavior. I'm still blaming the network. But they have new HMCs from ~4 months ago that is a .1 level back and hasn't experienced the issue. The new is..V11 R1 SP1111, previous is V11 R1 SP1110. I'm blaming network or HMC level, and despite stating as such, I really don't think its network.

Latest from support suggested this fix which clearly has a line item in it about GUI performance and task hanging.
https://delivery04.dhe.ibm.com/sar/CMA/HMA/0dq9e/1/MF71753.readme.html

Stay tuned.

José Pina Coelho
Thu April 02, 2026 11:21 AM

128GB to run an HMC is borderline insane.  IIRC, v11 has almost no difference from v10 (at least none that justifies requiring more than 64GB).

My homelab (a Thinkpad T460 with 32GB RAM), is running docker (with ollama, jellyfin, paperless, glpi, MQSeries, and portainer), plus a full GUI (GNOME, Firefox, Thunderbird, Eclipse), and it barely touches 16GB.

One thing that consistently messes responsiveness is (drum roll) DNS.  Machines that "don't use" DNS, are frequently asking themselves to resolve names and IP, then timing out on the non-existing response.  Another common problem mode are internal DNS servers that timeout on RFC1918 addresses (and the HMC has plenty of those).

Shawn Bodily
Thu April 02, 2026 09:49 AM

These manage ONLY the 3 new E1150s that have nothing configured yet, so not doing much of anything yet. I still blame the network.. :)

I believe customer has opened a support case, hence where the comment came from that they recommended 128GB for v11. Which I find nothing no documentation to support even as a recommendation. 

ED LANDMAN
Thu April 02, 2026 05:31 AM

Hello Shawn,

About the " Support says v11 code should run on HMCs with 128GB of memory"

I don't think there is a requirement for 128 GB. Do you have a case number?

If you have a lot partitions with connection problems for RMC Active checks, then you can run into perf issues. 

Please open a HMC software case at IBM and send in a pedbg so we can check the logs.

BR

ED 

Zaki Jääskeläinen
Thu April 02, 2026 03:38 AM

128GB RAM for the HMC. So, a small customer may end up with an HMC having more memory than the Power11 server it is managing. 🤭

It would be interesting to find out how that memory is utilized.