If I had to guess, it is a policy of the vendor and/or an oversight of the vendor. Most vendors purchase their IPMI firmware from a third party, then make their modifications to it for their own release. I've seen several motherboards that "support DCMI" (i.e. ipmi-dcmi reports that it is supported as you saw), but internally they never actually implemented anything related to DCMI. My assumption is that ipmi-dcmi showing support is lingering bits from the 3rd party firmware.
Supermicro did support DCMI in the past (https://serverfault.com/questions/389224/power-usage-via-ipmi-or-bios-or/389355#389355). Sorry that I don't have better info. If you bug Supermicro, maybe they can eventually give you what you want :P Al On Wed, Jul 25, 2018 at 1:48 AM, Vladimir Stegailov <stegai...@gmail.com> wrote: > Hi Al, > > thank you for the response! > > Do you think that this problem is a flaw in the DCMI implementation of the > current motherboard firmware? > In this case, we could hope it is to be corrected in future versions. > > Or is it rather a policy of the vendor? It seems so since they don’t > provide the raw command for PSU monitoring. > Are we to abandon plans for using freeipmi-based accounting in SLURM for > this cluster? > > Vlad > > > On Tue, Jul 24, 2018 at 10:37 PM, Albert Chu <ch...@llnl.gov> wrote: > >> Hi Vladimir, >> >> > What could be the reason for this problem with FreeIPMI inability to >> > read power statistics? >> >> Internally, I don't know how Supermicro is obtaining the data they have >> in their output. It would not be surprising if it were via an OEM >> specific extension. It would also not be surprising the motherboard >> does not support as much DCMI as they advertise. >> >> I have seen some motherboards that allow some mechanisms (such as DCMI) >> via LAN but not via in-band. Perhaps it would be worthwhile trying -- >> get-system-power-statistics over LAN? >> >> Al >> >> >> On Tue, 2018-07-24 at 12:44 +0300, Vladimir Stegailov wrote: >> > Hi! >> > >> > I have a question concerning FreeIPMI usage on the novel Supermicro >> > A+ >> > Server 1023US-TR4 with two AMD Epyc CPUs. >> > >> > The BMC firmware of the corresponding H11DSU-iN motherboard has been >> > upgraded to the latest available version (1.30). >> > >> > The FreeIPMI test shows that the power monitoring is available: >> > >> > *host1:~ # ipmi-dcmi --get-dcmi-capability-info* >> > *DCMI Specification Conformance : 1.1* >> > >> > *Power Management / Monitoring Support : Available* >> > *In-band System Interface Channel : Available* >> > *Serial TMODE : Not present* >> > *Out-Of-Band Secondary LAN Channel : Not present* >> > >> > *Number of SEL entries : 512* >> > *Record Level SEL Flush upon Rollover : Not present* >> > *Entire SEL Flush upon Rollover : Not present* >> > *SEL automatic rollover : Not present* >> > *DHCP Host Name : Not present* >> > *Asset Tag : Available* >> > >> > *Power Management Device Slave Address : 10h* >> > *Power Management Controller Device Revision : 1* >> > *Power Management Controller Channel Number : 0* >> > >> > *Primary LAN Out-of-band Channel Number : 1* >> > *Secondary LAN Out-of-band Channel Number : Not supported* >> > *Serial Out-of-band TMODE Capability Channel Number : Not supported* >> > >> > *ipmi_cmd_dcmi_get_dcmi_capability_info_enhanced_system_power_statist >> > ics_attributes: >> > bad completion code* >> > >> > However it is not possible to get power statistics by a DCMI request >> > (it >> > stops with an error): >> > >> > *host1:~ # ipmi-dcmi --debug --get-system-power-statistic* >> > *=====================================================* >> > *Group Extension - Get Power Reading Request* >> > *=====================================================* >> > *[ 2h] = cmd[ 8b]* >> > *[ DCh] = group_extension_identification[ 8b]* >> > *[ 1h] = mode[ 8b]* >> > *[ 0h] = mode_attributes[ 8b]* >> > *[ 0h] = reserved[ 8b]* >> > *=====================================================* >> > *Group Extension - Get Power Reading Response* >> > *=====================================================* >> > *[ 2h] = cmd[ 8b]* >> > *[ FFh] = comp_code[ 8b]* >> > *ipmi_cmd_dcmi_get_power_reading: bad completion code* >> > >> > At the same time, the Supermicro tool does show power statistics: >> > >> > *vvsteg@head-epyc:~/repos/SMCIPMITool_2.20.0_build.180525_bundleJRE_L >> > inux> >> > ./SMCIPMITool ipmi1 ADMIN ADMIN pminfo* >> > * [SlaveAddress = B0h] [Module 1]* >> > * Item | Value * >> > * ---- | ----- * >> > * Status | ON (00 00) * >> > * Input Voltage | 225.75 V * >> > * Input Current | 0.8 A * >> > * Main Output Voltage | 12.16 V * >> > * Main Output Current | 13.82 A * >> > * Temperature 1 | 31C/88F * >> > * Temperature 2 | 38C/100F * >> > * Fan 1 | 2624 RPM * >> > * Fan 2 | 0 RPM * >> > * Main Output Power | 168 W * >> > * Input Power | 176 W * >> > * PMBus Revision | 0x22 * >> > * PWS Serial Number | P1K0ACH21AN0368 * >> > * PWS Module Number | PWS-1K02A-1R * >> > * PWS Revision | 1.2 * >> > >> > What could be the reason for this problem with FreeIPMI inability to >> > read >> > power statistics? >> > (Unfortunately, we are not able to get any help from Supermicro with >> > this >> > issue.) >> > >> > Thank you. >> > >> > Vlad >> > _______________________________________________ >> > Freeipmi-users mailing list >> > Freeipmi-users@gnu.org >> > https://lists.gnu.org/mailman/listinfo/freeipmi-users >> -- >> Albert Chu >> ch...@llnl.gov >> Computer Scientist >> High Performance Systems Division >> Lawrence Livermore National Laboratory >> >> > _______________________________________________ > Freeipmi-users mailing list > Freeipmi-users@gnu.org > https://lists.gnu.org/mailman/listinfo/freeipmi-users _______________________________________________ Freeipmi-users mailing list Freeipmi-users@gnu.org https://lists.gnu.org/mailman/listinfo/freeipmi-users
