Although the issue was described here already before, after several days of runtime the noisy fan speed issue has returned. It is caused by the missing SDD temperature reported from the amsd client to the iLo interface. Just restarting amsd does not help. Manual fix is to reboot the whole XCP host and to reboot the iLO. But this is not feasible in production. Instead i found the following sequence of service reboots to resolve it. When the connectivity between amsd and ilo gets lost the amsd service remains in failed mode:
systemctl status amsd
amsd.service - Agentless Management Service daemon
Loaded: loaded (/usr/lib/systemd/system/amsd.service; enabled; vendor preset: enabled)
Active: failed (Result: timeout) since xxxx
The following bash script to be called as a cronjob (e.g. every hour)
#!/bin/bash
LOCKFILE="/run/check-amsd.lock"
(
flock -n 9 || exit 0
for svc in smad amsd cpqIde; do
if ! systemctl is-active --quiet "$svc"; then
hponcfg -b
sleep 300
systemctl reset-failed smad amsd cpqIde
systemctl restart smad
sleep 10
systemctl restart amsd
sleep 2
systemctl restart cpqIde
exit 0
fi
done
) 9>"$LOCKFILE"
For this to work you need to install both amsd and hponcfg from
https://downloads.linux.hpe.com/repo/mcp/XenServer/8/x86_64/current/
Note that this will cause a reboot of the ILO.
