CPLD firmware update for SY 660?

[DanCernese]

"...there is NOT a new version of CPLD firmware for the SY660..."

Yet.  

For the SY660 S/N, the issue is only in the range of 116 through 210.  to be published as soon as practical, but the logic is not the same and triage and validation cycles are still ongoing.  no forecast at this time.

[PatrickLong]

@DanCernese  So if I am understanding HPE recommendations correctly...  I should apply the 0x1c SY660 CPLD update from 2019 per Customer Advisory a00077281en_us to resolve the stuck-on-amber inconsequential mobo light on my SY 660's - regardless of their serial # designation - BUT I would only need to then subsequently apply this yet-to-be-released SY660 CPLD to SY660's in the serial # range 116 through 210 as the 4th through 6th digit per guidance.  None of my SY660's fall into that serial range, so I will NOT need to apply the new CPLD when it is released.  Do I have that correct?

Extra credit Dev assignment:  Can HPE please program logic into the CPLD binary that reads in the current mobo serial# and evaulates via regex to determine if it should be applied or not?  Particularly if it is strongly not recommended (or even potentially dangerous) to apply the forthcoming CPLD to an SY660 with a non-qualifying serial #... relying on human eyes to make this assessment is fraught with risk.

[PatrickLong]

I see that the CPLD firmware 0x1D for a subset of SY660 comute modules as well as a Customer Advisory a00121347en_us has now been released, and as with the SY480 CPLD release, this Advisory identifies very strict serial number range requirements for proper application of this CPLD firmware in the Advisory document's Scope section.  However, also like the SY48 release, this same serial number range restriction information is NOT included on either the Installation Instructions or the Release Notes tabs of the actual download page - this is an inconsistency that is very likely to cause confusion and misapplied updates along with whatever potential negative ramifications that would bring.  If this scope restriction to only a subset of serial numbers is in fact as important a requirement for the CPLD releases as the Customer Advisories for both the SY480 and SY660 would lead one to believe, I would ask that HPE please update those download pages with the same information for the sake of consistency.

[DanCernese]

fyi-- SY660 have 2 CPLD. Engineering discovered that early models of SY660 that were outside the affected SN range have a very early version of CPLD on one of the chips.. ..and that's the one that fails to flash correctly from that old version to the newest.  So we will emphasize NOT to apply outside the SN range for SY660.  In parallel we're researching if we can modify the package to check first; but that turns out not to be trivial for CPLD.

[PatrickLong]

@DanCernese any further info on this SY660 CPLD issue that you can divulge here?  I have an SY660 that is well outside of the S/N range 116 through 210 specified in the Advisory, (my 4th-6th digits are 842) yet this server has now failed twice in production with one of the failure modes listed in the Advisory "the system may power off abruptly during runtime and require an E-fuse reset to restart it."  Also same as the SY660 CPLD 0x1D download page Description tab  "1. A runtime shutdown with no standby power and no associated entry in the IML." which required an e-fuse reset to power on the server. Absolutely nothing identifiable in IML,  AHS, or ESXi  logs related to the source of these failures.

I had opened Support Case 5365360842 where they eventually tacitly advised me to upgrade the CPLD from current 0x1C to 0x1D:;  Support stated  "We do not see Fault Start / Fault Stop in AHS all event logs, however we have heard that there are servers which are out of that range which could possibly be affected.  If the system is up now, then go ahead and flash the CPLD FW.".  I'm just looking for your thoughts before I move forward and flash this thing, particularly given the May 9th, 22 addition to the Advisory "IMPORTANT: The CPLD version 0x1D must not be applied to HPE SY660 compute modules which are not in the affected serial number range. Doing so may prevent the compute module from powering on and require downgrading to an older CPLD version in order to recover." 

[DanCernese]

I am not aware of engineering having worked on this topic any further and I believe taking action not based on data from engineering is very risky..  I am unaware of any server outside the range being affected of any model unless the system board had been replaced.

[DanCernese]

I see they closed the case; please ask them to re-open it and elevate to engineering and point them to the advisory IMPORTANT note.  Given that the symptoms do not match, I would not apply the update unless you are prepared to potentially have to recover the server in the case where it does not resolve the issue.