Skip to content

Help: IPMI fan curves ignore my custom settings after reboot

Dedicated Servers by jane_ffm 3 replies 290 views
6 #1

Hi there! Thanks for reaching out about your IPMI fan curve issue. I'm sorry to hear your server is running loud — that definitely isn't the experience we want for our Contabo customers.

A few things I'd like to check with you:

• Which chassis model are you using? (e.g., our FM-2010, FM-2020, or FM-4020 platform)
• What BMC firmware version is currently installed?
• Are you seeing this behavior on a fresh boot, or only after a warm reboot?

In some cases, we've observed that custom fan curves stored in the BMC NVRAM can be overwritten by the default thermal profile during certain boot sequences. I'd be happy to check our internal knowledge base for any known firmware quirks matching your setup. Feel free to DM me your service tag and I can verify whether you're on a revision with confirmed fixes.

— Jane @ Contabo

Single mode till I die 💀
2 #2

I have seen a man lose a raid array because his fan curve reset script failed silently for three weeks. The drives cooked. The backups were unreadable. Verify your restores.

I run ipmitool raw commands every 5 minutes on my Leaseweb box — https://www.leaseweb.com. It is not elegant. It is not pretty. But my drives are 34C and my sanity is intact. The firmware bug is real. The workaround is realer. Do not trust bmc vendors to fix anything ever.

4 #3

- confirmed behavior: custom ipmi fan curves do not persist across reboot on multiple supermicro boards with bmc firmware 3.88 and below.
- root cause: firmware writes default thermal profile over user nvram on power-on self-test.
- workaround: cron job executing ipmitool raw 0x30 0x70 0x66 0x01 0x00 0x64 every 300 seconds.
- production deployment: twenty-three months without thermal event.
- additional notes:
- bmc watchdog does not catch this.
- ipmi sel logs no error.
- vendor acknowledged bug in private ticket; no fix timeline provided.

-t

worst bandwidth, best stories
#4

The someoneelsesFirmwareBug is always the someoneelsesProblem until it is YOUR Problem at 3AM with a screaming InterServer server and a NOC that does not answer the phone[1]. I have run this cronjob in Production for two years now on three different KnownHost boxes and the FanCurves still reset on every Reboot but the cronjob fixes it before the drives hit 50C. The BMC is a Computer inside your Computer that you cannot Control and they want you to forget this fact. The IPMI specification is a Joke that the vendors do not read. I asked CloudCone for a firmware update and they sent me a PDF from 2019. This is the State of the Industry. We are all just cronjobs holding together someoneelsesGarbage.

[1] private ticket #18473, "acknowledged, no eta, kindly wait" — the Kindly is how you know they will never fix it.

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft