Cray has identified an issue with ipmitool LAN commands on Jefferson Pass, Kennedy Pass and Wildcat Pass Intel platforms. An error is received following an ipmitool power command over the LAN interface and has been reproduced across all BIOS/BMC firmware releases. Example:
Errors received are:
Close Session command failed
Error: Unable to establish IPMI v2 / RMCP+ session
Unable to set Chassis Power Control to Cycle
Note that when these commands are run locally, no error is observed. Additionally, when a power cycle is run remotely over LAN, the remote server does actually respond to the request and performs the cycle – there appears to be a timing issue on the sender’s end in that it does not receive the response from the remote server. Intel has stated that there appears to be a race condition for the BMC after the reset command is issued, and that a possible solution might be to issue a pause into the power cycle command to give it time to finish the close session command. Please let us know if there is any recourse for us to work towards a resolution.
Thanks,
Ben Weisnicht
Hello,
I've seen your mail in mailing list, but I have questions rather than solutions.
If it's a bug in BMC, shouldn't be Intel fixing it in BMC? Also, why isn't Intel working at the fix be it in BMC or ipmitool? They certainly have manpower to do so, but the most importantly they have (developer) access to affected platforms.
I'm also not clear whether you're reporting a bug or asking for advice how to fix this.
I'm sorry for kind of jerky reply and thank you for clarification.
Best regards,
Z.