Describe the enhancement:
Report Linux Pressure Stall Information (PSI) metrics in the beat metrics. Include PSI info when
Describe a specific use case for the enhancement or feature:
When reviewing diagnostic snapshots this information would be used to detect if CPU, Memory, or IO pressure could be causing processing to stall. If this is occurring a lot (i.e. high percentage over avg300) then this would be strong signal that some aspect of the host is overloaded. And that might be a cause of the reported issue.
Here's are some example of reading PSI info from the CLI.
System Level PSI
cat /proc/pressure/{cpu,io,irq,memory}
some avg10=0.09 avg60=0.23 avg300=0.34 total=29210385599
full avg10=0.00 avg60=0.00 avg300=0.00 total=0
some avg10=0.00 avg60=0.00 avg300=0.00 total=2022562678
full avg10=0.00 avg60=0.00 avg300=0.00 total=1843728046
full avg10=0.00 avg60=0.00 avg300=0.00 total=16487083525
some avg10=0.00 avg60=0.00 avg300=0.00 total=2107930
full avg10=0.00 avg60=0.00 avg300=0.00 total=1986722
Cgroup V2 PSI
If the host is using cgroup v2 and the process is a member of a cgroup, then we can get PSI information scoped to the tasks in the group.
cat /etc/mtab | grep cgroup
cgroup /sys/fs/cgroup cgroup2 ro,nosuid,nodev,noexec,relatime,nsdelegate,memory_recursiveprot 0 0
ecf21fadb472:~# cat /proc/self/cgroup
0::/
cat /sys/fs/cgroup/{cpu,io,irq,memory}.pressure
some avg10=0.00 avg60=0.00 avg300=0.00 total=41743102
full avg10=0.00 avg60=0.00 avg300=0.00 total=40889871
some avg10=0.01 avg60=0.09 avg300=0.03 total=63542837
full avg10=0.01 avg60=0.09 avg300=0.03 total=63432991
full avg10=0.00 avg60=0.00 avg300=0.00 total=64963754
some avg10=0.00 avg60=0.00 avg300=0.00 total=0
full avg10=0.00 avg60=0.00 avg300=0.00 total=0
Describe the enhancement:
Report Linux Pressure Stall Information (PSI) metrics in the beat metrics. Include PSI info when
Describe a specific use case for the enhancement or feature:
When reviewing diagnostic snapshots this information would be used to detect if CPU, Memory, or IO pressure could be causing processing to stall. If this is occurring a lot (i.e. high percentage over avg300) then this would be strong signal that some aspect of the host is overloaded. And that might be a cause of the reported issue.
Here's are some example of reading PSI info from the CLI.
System Level PSI
Cgroup V2 PSI
If the host is using cgroup v2 and the process is a member of a cgroup, then we can get PSI information scoped to the tasks in the group.