useful sar counters

gistfile1.txt

# canonical sar command invocation
sar -f [FILE] [flags] [interval] [count]

# general purpose flags
# -f [FILE] refers to a file in /var/log/sa/saXX where XX is the day of month
# -o [FILE] capture all performance data in a binary file to be read with sar later
#           Note: *all* performance counters are captured by default
# -p flag will pretty print device names rather than using dev{m}-{n} format
# when [interval] is 0 sar will report statistics for time since boot
# when [count] is omitted sar will loop indefinitely

# table of sar flags along with some of the most useful counters
    -u      cpu utilization
            %user     : % CPU utilization in user land
            %system   : % CPU utilization in kernel land
            %iowait   : % CPU idle during which system waiting on IO
            %idle     : % CPU idle during which system was *not* waiting on IO
            %steal    : % of time system waiting while hypervisor servicing another tenant
    -q      run queue & load average
            runq-sz   : number of tasks waiting for run time
            ldavg-(X) : system load average over X interval
    -B      paging statistics
            fault/s   : major & minor faults / sec
            majflt/s  : major faults / sec (require read from disk)
    -H      huge pages
    -r      memory utilization
            %memused  : % of used memory
            %commit   : % of memory needed by workload relative to RAM+swap
    -R      memory statistics
    -S      swap utilization
    -W      swap statistics
    -v      kernel tables
    -d      disk statistics per block device
            tps       : number of read & write ops / sec
            rd_sec/s  : number of 512b sectors read / sec 
            wr_sec/s  : number of 512b sectors write / sec
            await     : average wait + service time (ms)
            svctm     : average service time (ms)
            %util     : % of cpu time during which device busy
    -b      io rates to all physical devices
            tps       : number of read & write ops / second
            rtps      : read ops / second
            wtps      : write ops / second
    -w      task creation & context switch
            cswch/s   : number of context switches / sec

# examples of flags

# display run queue length and load average
sar -f [FILE] -q
sar -q [interval] [count]

# display memory utilization counters
sar -f [FILE] -r
sar -r [interval] [count]

# display all cpu utilization counters
sar -u ALL

# display per-cpu utilization counter
sar -P [0|1|...|ALL]

# display disk statistics with pretty printed device names
sar -dp [interval] [count]

# display per-device (-n) counter

# table of supported arguments to (-n)
    DEV     nic stats
    EDEV    nic errors
    IP      ip stats
    EIP     ip errors
    TCP     tcp stats
    ETCP    tcp errors
    SOCK    socket stats

# display number of ipv4 tcp and udp sockets open
sar -f [FILE] -n SOCK

# use LC_TIME environment variable to have sar print dates in 24h time
env LC_TIME=POSIX sar -u 1

# use S_TIME_FORMAT environment variable to have sar print date in YYYY-MM-DD format
env S_TIME_FORMAT=ISO sar -u 1