# derived metrics for processes # metrics to provide additional process details proc.psinfo.age = kernel.all.uptime - proc.psinfo.start_time proc.psinfo.age(oneline) = time in seconds since process was started proc.psinfo.age(helptext) = '\ time in seconds since process was started, calculated from proc.psinfo.start_time subtracted from kernel.all.uptime.' proc.io.total_bytes = instant(proc.io.read_bytes) + instant(proc.io.write_bytes) - instant(proc.io.cancelled_write_bytes) proc.io.total_bytes(oneline) = total bytes read and written by process proc.io.total_bytes(helptext) = '\ total bytes read and written by the process, less cancelled written bytes.' # metrics to identify resource hogs proc.hog.cpu = 100 * (rate(proc.psinfo.utime) + rate(proc.psinfo.stime)) / (kernel.all.uptime - proc.psinfo.start_time) proc.hog.cpu(oneline) = average percentage CPU utilization of each process proc.hog.cpu(helptext) = '\ average CPU utilization of each process expressed as a percentage of time since the process started.' proc.hog.mem = proc.memory.vmrss + proc.memory.vmswap proc.hog.mem(oneline) = sum of resident and swapped memory used by the process proc.hog.mem(helptext) = '\ amount of resident and swapped memory used by each process.' proc.hog.disk = (rate(proc.io.read_bytes) + rate(proc.io.write_bytes) - rate(proc.io.cancelled_write_bytes)) / (kernel.all.uptime - proc.psinfo.start_time) proc.hog.disk(oneline) = average I/O rate of each process since it was started proc.hog.disk(helptext) = '\ average I/O rate (reads and writes less cancelled writes) of each process since it was started.' proc.hog.net = rate(bpf.proc.net.tcp.recv.bytes) + rate(bpf.proc.net.tcp.send.bytes) + rate(bpf.proc.net.udp.recv.bytes) + rate(bpf.proc.net.udp.send.bytes) proc.hog.net(oneline) = sum of network usage by the process proc.hog.net(helptext) = '\ sum of network usage (TCP/UDP recv/send) by each process.'