[SOLVED] Wonky IPMI on SM X9SCM-iiF

tic226

New Member
Feb 2, 2017
20
3
3
119
DE
Last week i bought a used X9SCM-iiF on ebay, threw in a E3-1225v2 (according to SM the E3-xxx5 variants with integrated graphics are not 'recommended', for now i assume that doesn't cause the IPMI issue..) and 2x8GB DDR3. It's working fine except the IPMI is causing trouble;

The IPMI interface wasn't pulling an IP via DHCP on its dedicated NIC, so i've set a static one trough BIOS and rebooted (complete power off, i read a lot of horror stories about SM IPMI..) but couldn't ping the address, BIOS says IPMI is 'working' and shows the correct settings. I checked the other 2 ports as well, nothing, so i installed Win7 on the machine to be able to use the ipmitools from SM. I found IPMICFG and used -fed to reset to factory defaults, rebooted and tried configuring it from scratch to no avail, still no ping to the set address.

Next stop: firmware updates. I got the latest IPMI firmware and while i was already at it, updated the BIOS as well with the FreeDOS bootdisk method. Went fine, still the same issues. So i started suspecting my network, specifically my switch (1820-24G). The port i connected the IPMI NIC to was configured for VLAN1, untagged at that time. I tried setting VLAN tag through IPMICFG, different settings on the switch port (tagged/untagged, even in non-sensical combinations) and after nothing helped tried different switch ports. Still nothing.

Out of pure frustration (and after contemplating ditching IPMI on that machine alltogether) i put a live ubuntu disk in my notebook (to have nothing VLAN related in the network configs) and connected that directly to the IPMI and BAM, link. I proceeded to open the webGUI, set the IPMI NIC config to 'dedicated', rebooted (via power off) and connected the IPMI NIC to an untagged port on the switch again: again no connection. The link LED on the jack is lit alright.

I tried disabling jumbo frames on the switch which is (apart from VLAN) the only setting that sets this switch apart from a 'dumb' switch. Do you have any idea in which direction i should look or might this be a busted IPMI? Maybe for some reason it just hates my switch?

FWIW, after the IPMI firmware update the SuperDoctor program (as well as the IPMI webGUI) didn't show the CPU temp anymore, only 'LOW' and 'MEDIUM' and the other 2 temps are both displayed as 28°C and 34°C statically which must be bogus, during a Prime95 test the CPU went up to 78°C (transitioned from 'LOW' to 'MEDIUM' which i would consider 'HIGH'..) and the exhaust air was quite warm , the other system temps must have changed during that time but always show 28/34. Memtest86 shows plausible CPU temps though so the on-die sensors must be fine. I'm quite certain the CPU temp value was displayed as °C before the firmware update.
 

Blinky 42

Active Member
Aug 6, 2015
563
201
43
44
PA, USA
I have not see that level of trouble since the old x7 series boards with add-on cards for ipmi.
Have you done a full power cycle after switching it to be dedicated port for IPMI ? You can also do ipmicfg -r to have it reboot the ipmi controller itself.
Just a software / power button power cycle doesn't reset the ipmi side at all.
 

tic226

New Member
Feb 2, 2017
20
3
3
119
DE
I rebooted with ipmicfg -r, later i power cycled just to be 'extra' sure. But i made some progress:
To confirm my suspicions about the switch i took a cheap 'dumb' switch, connected the IPMI NIC to that
switch and a second cable to the 1820-24G. It was a desperate move but it somehow works. I activated
VLAN tagging on the IPMI NIC, set the port on the 1820 to tagged, tried again and it still works. This confirms that
'actually' there's nothing wrong with the IPMI and even the VLAN tagging works. But why?

I dumbed down the 1820 as far as i could, no stp, jumbo frames,IGMP snooping,etc., hell i even restarted
the switch. Any ideas?
 

tic226

New Member
Feb 2, 2017
20
3
3
119
DE
We're you using the same cable each time? Possible bad Ethernet cable?
No, i tried different cables. It's working with the dumb switch inbetween. To be absolutely sure it's not a bad ethernet jack or cable i switched back and forth [directly connected <> switch in the middle] multiple times; different switch ports, different cables. It always worked with the switch in the middle but never directly connected to the 1820.
 
  • Like
Reactions: PigLover

tic226

New Member
Feb 2, 2017
20
3
3
119
DE
I found the culprit:
The 1820 switch has two 'Green Features' which i totally forgot about. They were both active and after deactivating 'Low Power Idle' i could connect to the IPMI NIC without issue. Only downside is the switch's power consumption jumped from 4W to 7W.

Although it's technically solved i still can't let go of it. I highly suspect the IPMI PHY (Realtek RTL8201N), the datasheet states a few supported energy saving techniques but still. At least i can now shoot the SM support a mail and ask about reported issues with energy saving settings on switches.
 
  • Like
Reactions: PigLover and Cole

tic226

New Member
Feb 2, 2017
20
3
3
119
DE
At-least it's only 3W too ;)
:D
It's complaining on a high level but still, the power draw almost doubled. I pay EUR0.27/kWh here in DE, every watt counts.
If it can't be completely solved in the end i think i will only deactivate low power idle on the switch if i really have to use iKVM/virtual media and that doesn't happen too often. The IPMI doesn't support SNMP and i use observium for SNMP monitoring which supports IPMI polling but doesn't let me add non-SNMP devices to begin with.
 

frogtech

Well-Known Member
Jan 4, 2016
1,336
208
63
32
I had IPMI/BMC issues on an X8 motherboard when I had the SuperDoctor installed. In fact SM support told me to uninstall it as it can cause conflicts with the IPMI.