// pmieconf-pmie 1 /var/lib/pcp/config/pmieconf
// 1 dmthin.data_high_util enabled = yes
// 1 dmthin.metadata_high_util enabled = yes
// 1 primary.pmda_restart enabled = yes
// end
//
// --- START GENERATED SECTION (do not change this section) ---
// Auto-generated by pmieconf on:  Mon Nov  4 14:12:53 2024
//

// 1 cpu.context_switch
delta = 2 min;
cpu.context_switch = 
some_host (
    kernel.all.pswitch 
	> hinv.ncpu  * 4000 count/sec
) -> syslog 10 min "High aggregate context switch rate" " %vctxsw/s@%h";

// 1 cpu.load_average
delta = 2 min;
cpu.load_average = 
some_host (
    // threshold scales with the number of CPUs (works better for
    // large systems) and there is an absolute lower bound,
    // especially for small systems
    kernel.all.load  #'1 minute' > hinv.ncpu  * 3
    && kernel.all.load  #'1 minute' > 4
) -> syslog 10 min "High 1-minute load average" " %vload@%h";

// 1 cpu.system
delta = 2 min;
cpu.system = 
some_host (
    // first term is always true, but provides %v for actions ...
    ( 100 * kernel.all.cpu.sys  / hinv.ncpu  ) > 0
    && 100 * ( kernel.all.cpu.user  + kernel.all.cpu.sys  )
	    > 70 * hinv.ncpu 
    && 100 * kernel.all.cpu.sys  /
	    ( kernel.all.cpu.user  + kernel.all.cpu.sys  )
		> 75
) -> syslog 10 min "Busy executing in system mode" " %v%sys@%h";

// 1 cpu.util
delta = 2 min;
cpu.util = 
some_host (
    100 * ( kernel.all.cpu.user  +
	    kernel.all.cpu.sys  +
	    kernel.all.cpu.intr  ) / hinv.ncpu 
	> 90
) -> syslog 10 min "High average processor utilization" " %v%util@%h";

// 1 dmthin.data_high_util
delta = 10 mins;
dmthin.data_high_util = 
some_host (
    some_inst (
        ( 100 * dmthin.pool.data.used  /
                dmthin.pool.data.total  )
        > 95
    )
) -> syslog 10 min "dmthin pool data is filling up" " %v%used [%i]@%h";

// 1 dmthin.metadata_high_util
delta = 10 mins;
dmthin.metadata_high_util = 
some_host (
    some_inst (
        ( 100 * dmthin.pool.metadata.used  /
                dmthin.pool.metadata.total  )
        > 80
    )
) -> syslog 10 min "dmthin pool metadata is filling up" " %v%used [%i]@%h";

// 1 entropy.available
delta = 2 min;
entropy.available = 
some_host (
	( 100 * kernel.all.entropy.avail  /
          kernel.all.entropy.poolsize  ) >= 0 &&
	( kernel.all.entropy.avail <= 150 )
) -> syslog 10 min "Low random number entropy available" " %v%avail@%h";

// 1 filesys.filling
delta = 4 mins;
filesys.filling = 
some_host (
    some_inst (
	( 100 * filesys.used  / filesys.capacity  ) > 95
	&& filesys.used  + 
	    20 min * ( rate filesys.used  ) >
		filesys.capacity 
    )
) -> syslog 10 min "File system is filling up" " %v%used[%i]@%h";

// 1 memory.exhausted
delta = 2 min;
memory.exhausted = 
some_host (
    ( avg_sample ( swap.pagesout  @0..9 ) ) > 0 &&
    30 %_sample swap.pagesout  @0..9 >= 5
) -> syslog 10 min "Severe demand for real memory" " %vpgsout/s@%h";

// 1 memory.oom_kill
delta = 2 min;
memory.oom_kill = 
some_host (
        mem.vmstat.oom_kill  > 0
) -> syslog 10 min "OOM kill" " %v/sec@%h";

// 1 mssql.stolen_node_memory
delta = 2 min;
mssql.stolen_node_memory = 
some_host (
    ( avg_sample ( mssql.memory_manager.stolen_node_memory  @0..9 ) ) > 0
    && 60 %_sample
    ( 100 * ( mssql.memory_manager.stolen_node_memory @0..9  /
              mssql.memory_manager.total_server_memory @0..9  ) )
	> 70
) -> syslog 10 min "High mssql stolen node memory" " %v%mem@%h";

// 1 network.listenoverflows
delta = 2 min;
network.listenoverflows = 
some_host (
        network.tcp.listenoverflows  > 0
) -> syslog 10 min "TCP listen overflows" " %v/sec@%h";

// 1 network.tcpqfulldocookies
delta = 2 min;
network.tcpqfulldocookies = 
some_host (
        network.tcp.reqqfulldocookies  > 0
) -> syslog 10 min "TCP request queue full SYN cookie replies" " %v/sec@%h";

// 1 network.tcpqfulldrops
delta = 2 min;
network.tcpqfulldrops = 
some_host (
        network.tcp.reqqfulldrop  > 0
) -> syslog 10 min "TCP request queue full drops" " %v/sec@%h";

// 1 per_cpu.many_util
delta = 2 min;
per_cpu.many_util = 
some_host (
    hinv.ncpu  > 12 &&
    80 %_inst (
	100 * ( kernel.percpu.cpu.user  +
		kernel.percpu.cpu.sys  +
		kernel.percpu.cpu.intr  )
	    > 95
    )
) -> syslog 10 min "High number of saturated processors" " >80%CPUs@%h";

// 1 per_cpu.some_util
delta = 2 min;
per_cpu.some_util = 
some_host (
    some_inst (
      ( 100 * ( kernel.percpu.cpu.user  +
		kernel.percpu.cpu.sys  +
		kernel.percpu.cpu.intr  ) )
	    > 95
    )
    && hinv.ncpu  > 1
    && hinv.ncpu  <= 12
) -> syslog 10 min "High per CPU processor utilization" " %v%util[%i]@%h";

// 1 per_cpu.system
delta = 2 min;
per_cpu.system = 
some_host (
    some_inst (
	// first term is always true, but provides %v for actions ...
	( 100 * kernel.percpu.cpu.sys  ) > 0
	&& 100 * ( kernel.percpu.cpu.user  +
		   kernel.percpu.cpu.sys  ) > 75
	&& 100 * kernel.percpu.cpu.sys  /
	    ( kernel.percpu.cpu.user  + kernel.percpu.cpu.sys  )
		> 80
    )
    && hinv.ncpu  > 1
) -> syslog 10 min "Some CPU busy executing in system mode" " %v%sys[%i]@%h";

// 1 per_disk.average_queue_length
delta = 2 min;
per_disk.average_queue_length = 
some_host (
    some_inst (
      ( disk.dev.aveq  @0 > disk.dev.aveq  @1 )
        &&
      ( disk.dev.avactive  * 100 > 95 )
    )
) -> syslog 10 min "High per disk average queue length" " %vaveq[%i]@%h";

// 1 per_disk.average_wait_time
delta = 2 min;
per_disk.average_wait_time = 
some_host (
  some_inst (
    ( disk.dev.total_rawactive  @0 / disk.dev.total  @0 ) * 100
      >
    ( disk.dev.total_rawactive  @1 / disk.dev.total  @1 ) * 100
      &&
    ( disk.dev.avactive  * 100 > 95 )
  )
) -> syslog 10 min "High per disk average queue length" " %v%await[%i]@%h";

// 1 per_netif.collisions
delta = 2 min;
per_netif.collisions = 
some_inst (
    100 * network.interface.collisions 
	    / ( network.interface.collisions 
		+ network.interface.out.packets  )
	> 40 &&
    network.interface.out.packets 
	> 10 count/second
) -> syslog 10 min "High collision rate in packet sends" " %v%collisions[%i]@%h";

// 1 per_netif.errors
delta = 2 min;
per_netif.errors = 
some_inst (
    network.interface.total.errors  
	> 15
) -> syslog 10 min "High network interface error rate" " %verr/s[%i]@%h";

// 1 per_netif.util
delta = 2 min;
per_netif.util = 
some_inst (
    ( 100 * network.interface.total.bytes   /
	    network.interface.baudrate   )
	       > 85
    && network.interface.baudrate   > 0
) -> syslog 10 min "High network interface utilization" " %v%util[%i]@%h";

// 1 power.thermal_throttle
delta = 2 min;
power.thermal_throttle = 
some_host (
	100 * hinv.cpu.thermal_throttle.package.time  > 50
) -> syslog 10 min "CPU is experiencing thermal throttling" " %v%time[%i]@%h";

// 1 primary.pmda_restart
delta = 2 min;
primary.pmda_restart = 
some_inst (
    pmcd.agent.status != 0
) -> shell 10 min "pmsignal -s HUP -p pmcd"
  & syslog 10 min "Restart unresponsive PMDAs" " pmda%i[%v]@%h";

// --- END GENERATED SECTION (changes below will be preserved) ---