Yeah, same, better Experience with Aliexpress than eBay.
Inspected or not, since I was using a Freight Forwarding Service (via Sea Shipping, so > 90 Days anyways), eBay Buyer Protection doesn't apply. It was too late for it anyways.
EXPIRED HYVE Edge Metal G10 - Epyc 7642 - 250£/offer
- Thread starter Cruzader
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Inspected or not, since I was using a Freight Forwarding Service (via Sea Shipping, so > 90 Days anyways), eBay Buyer Protection doesn't apply. It was too late for it anyways.
"Only" after Page 7.What happened in that thread? Seems that all messages got deleted, except @luckylinux messages.
@Patrick: seems like (almost ?) my Posts got restored, but apparently other disappeared ? Could it be that re-inserting Posts overwrote some unique IDs in the Database for instance ?
Yeah, I had the most direct feedback. Account blocked on Monday Morning (European Time). Good Way to wake up for the Day !
Also @Patrick had it rough trying to restore my Account and Posts
.In case someone wants a printable NVME mount/fastener for the motherboard, just twist it in to fasten the disk:
Group Question - My metal edge server M2 TEMP sensor has always thrown a warning and reported the temp to be 0.037degrees, which makes the fans "hunt". Has anyone else seen this issue? I can't even find a diagram to tell me where to look to see if this sensor is damaged.
Any insight or help is much appreciated.
Any insight or help is much appreciated.
I manage to control the fan speed and it seems possible to remove sensors (which might help @Ulgracht).
You need to set a user/password for OpenBMC and enable DHCP. Then, you can SSH into OpenBMC. To be clear: instead of SSH to the server itself, you SSH to the IP attached to OpenBMC (the IP attach to the BMC ethernet port).
The SSH credentials is the same user/password of OpenBMC (which you use in the web browser interface).
Then, you can read and modify one file (`/usr/share/swampd/config.json` and can create a copy to `/etc/swampd/config.json`). That file seems to be responsible to control the fan-speed, based on each temperature sensor. MAYBE you can remove the damaged sensor (not a fix, but mitigates the issue) or you can remove the low temperature warning.
In general:
Based on GPT/Claude/Gemini (...), you can also set a specific fan speed, using something like:
------
I didn't investigate it further.
I increase some "setpoints" to make the server accept higher temperature... Of course, it will run hotter and the fan will eventually get louder. My issue is "OCP NIC TEMP", I need to somehow improve the cooling of this NIC [help wanted]. Because that is what is causing the fans to go crazy (it's already the max of 95ºC, which I set as target).
However, as a experiment, it worked. Now, we need some smart way to control it AND make the NIC run cooler overall.
You need to set a user/password for OpenBMC and enable DHCP. Then, you can SSH into OpenBMC. To be clear: instead of SSH to the server itself, you SSH to the IP attached to OpenBMC (the IP attach to the BMC ethernet port).
The SSH credentials is the same user/password of OpenBMC (which you use in the web browser interface).
Then, you can read and modify one file (`/usr/share/swampd/config.json` and can create a copy to `/etc/swampd/config.json`). That file seems to be responsible to control the fan-speed, based on each temperature sensor. MAYBE you can remove the damaged sensor (not a fix, but mitigates the issue) or you can remove the low temperature warning.
In general:
Code:
ssh admin@{open-bmc-ip}
# Change the "setpoints" where it says "temps" (which is CPU, M2, OCP NIC...)
sudo vi /usr/share/swampd/config.json
# Make a copy (not sure if necessary, because I didn't restart)
sudo cp /usr/share/swampd/config.json /etc/swampd/config.json 2>/dev/null || true
# Restart the service
sudo systemctl restart phosphor-pid-control.servicebusctl set-property xyz.openbmc_project.FanSensor /xyz/openbmc_project/control/fanpwm/Pwm_1 xyz.openbmc_project.Control.FanPWM Target t 50. I don't think it will work, but it's like to have a way to set it.------
I didn't investigate it further.
I increase some "setpoints" to make the server accept higher temperature... Of course, it will run hotter and the fan will eventually get louder. My issue is "OCP NIC TEMP", I need to somehow improve the cooling of this NIC [help wanted]. Because that is what is causing the fans to go crazy (it's already the max of 95ºC, which I set as target).
However, as a experiment, it worked. Now, we need some smart way to control it AND make the NIC run cooler overall.
Last edited:
Well, I didn't test that yet, but I appreciated the Instructions.
If you (for whatever Reasons ... Ear Pain being one of them during low-Load testing with one Server stuck with Fans at full Speed for some weird Reason) want to reduce Fan Speeds to a "reasonable" fixed Level:
Code:
MYIPMIHOST:~$ sudo /bin/bash
MYIPMIHOST:/home/admin# systemctl stop phosphor-pid-control.service
MYIPMIHOST:/home/admin# sudo echo 160 > /sys/class/hwmon/hwmon0/pwm1
MYIPMIHOST:/home/admin# sudo echo 160 > /sys/class/hwmon/hwmon0/pwm2
MYIPMIHOST:/home/admin# sudo echo 160 > /sys/class/hwmon/hwmon0/pwm3
MYIPMIHOST:/home/admin# sudo echo 160 > /sys/class/hwmon/hwmon0/pwm4
MYIPMIHOST:/home/admin# sudo echo 160 > /sys/class/hwmon/hwmon0/pwm5
MYIPMIHOST:/home/admin# sudo echo 160 > /sys/class/hwmon/hwmon0/pwm6It always ends up into some Black Screen (there is no POST at all), unless I set a Boot Override in
Operations -> Server power operations -> Boot settings -> Boot settings overrideThe Fans stuck at full speed may have occurred only after I installed 2 x U.2 NVMe Drives into the Server, not entirely sure, as I don't recall that having occurred before. Normally, if "locked out", all I have to do is remove AC Power, let it rest say 30 Seconds, plug AC Power back in, then let it do its 2 Boots/Reboot Loops and finally it will load correctly. Not anymore. Unless I set something for
Boot setting override, then it just shows black Screen in KVM and it's stuck .I manually set the RAM speed to the default Setting via
Overclock -> Enabled and then Memory Clock Speed -> 1333MHz (I have 2666MHz RAM, divided by 2 since DDR).Not sure what else that I changed from default it might not have liked
.On another Note, I feel like the CR2032 RTC Battery is draining VERY quickly if the Server is not plugged in (after a Year the CR2032 Battery was reading 2.7 VDC in IPMI). Anybody else observing the same ?
Last edited:
I have been battling with one of mine which seemed to be screaming all the time under load and it has led me to some info which might be useful.
Firstly I assumed it might be the CPU running close to the high limit, so I removed the heatsink and replaced the crusty thermal paste, I also removed the OCP NIC as it was also hot AND I wasnt using it, immediately the CPU temperatures dropped but now half of the fans were running at 86%, the OCP NIC temperature was no longer visible in the sensors list via the BMC gui.
Since I had the NIC on the bench, I removed the heatsink and the thermal paste had turned into a hard grout, i had to chip it off of the processor and heatsink... So I put some new paste on, reinstalled it and now its running at ~50% PWM on the fans, cpu temp is high again but the fans are much quieter.
If it starts getting hot on the NIC again I might try putting something in to guide the air toward it.
Word of warning though, I had removed this NIC on other servers and the white plastic holding the heatsink in place had become brittle and they fell off as I removed the NIC from the board, indicating that they had been getting hot for some time.#
TLDR; just removing the NIC makes the fans ramp up as the sensor goes missing. and the thermal paste is probably going to need replacing.