A generated or hand-made config.pmlogger file may have many
clauses such as:
#+ kernel/summary-linux:y:default:
## summary kernel performance data [Linux]
log advisory on default {
mem.util
swap.pagesin
swap.pagesout
swap.free
swap.used
kernel.all.pswitch
kernel.all.intr
kernel.all.load
kernel.all.runnable
network.interface.collisions
network.interface.in.drops
network.interface.out.drops
disk.dev.avactive
filesys.full
}
[...]
#+ tools/collectl:y:default:
## metrics used by the collectl command
log advisory on default {
[...]
kernel.all.intr
kernel.all.load
kernel.all.nprocs
kernel.all.pswitch
kernel.all.runnable
[...]
#+ tools/pmstat:y:default:
## metrics used by the pmstat(1) command
log advisory on default {
kernel.all.load
swap.used
[...]
Note e.g. kernel.all.load being listed frequently, all with the "default"
logging interval.
pmlogger appears to interpret this as a requirement to log these
multiply-mentioned metrics multiple times, thus multiplying the
disk consumption. One can confirm this with e.g. pmlogger -r
outputs, or pmval -U $archive listing the same metric repeated
at near-identical timestamps, then the expected interval.
Instead, pmlogger should duplicate-eliminate amongst specification
groups.