Use this table on the supplied values:

Evidence in Prometheus	                                                                        Bottleneck
CPU utilisation above 85%, node_load1 above core count, iowait low	                            CPU
MemAvailable under 10% of total, swap in/out rising (node_vmstat_pswpin, node_vmstat_pswpout)	Memory
iowait mode high (above 20%), disk busy % near 100, CPU mostly idle	                            Disk I/O
rate(node_network_receive_bytes_total) or transmit near link speed, CPU and disk low	        Network
node_filesystem_avail_bytes under 10%	                                                        Disk space

# queries to use for the analysis
100 - (avg(rate(node_cpu_seconds_total{mode="idle"}[1m])) * 100)
avg(rate(node_cpu_seconds_total{mode="iowait"}[1m])) * 100
node_load1 / count(count by (cpu) (node_cpu_seconds_total))
node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes * 100
rate(node_vmstat_pswpout[1m])
rate(node_disk_io_time_seconds_total[1m]) * 100
sum(rate(node_network_receive_bytes_total{device!="lo"}[1m]))

# answer format:
# "Evidence: CPU 97%, load 6.2 on 4 cores, iowait 1%. Bottleneck: CPU.
#  Reason: run queue exceeds core count while disk and memory are healthy."