SymbX/systemd
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

report: report user and system CPU time per cgroup

Browse Source 
Extend io.systemd.CGroup.CpuUsage from a single per-unit nanosecond
counter to three rows distinguished by a "type" field of "total",
"user", or "system". The values come from cpu.stat's usage_usec,
user_usec and system_usec keys, read in a single keyed-attribute
fetch and cached on each CGroupInfo so each scrape only opens
cpu.stat once per cgroup.
This commit is contained in:
Yaping Li
2026-04-30 20:08:39 +02:00
committed by Zbigniew Jędrzejewski-Szmek
parent 4ce8cd17da
commit f78ba86d26
2 changed files with 105 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/units/TEST-74-AUX-UTILS.report.sh
+7
View File
@@ -37,6 +37,13 @@ varlinkctl list-methods /run/systemd/report/io.systemd.CGroup
varlinkctl --more call /run/systemd/report/io.systemd.CGroup io.systemd.Metrics.List {}
varlinkctl --more call /run/systemd/report/io.systemd.CGroup io.systemd.Metrics.Describe {}
# CpuUsage emits one row per (cgroup, type) where type is total, user, or system.
# Confirm all three are present.
cgroup_metrics=$(varlinkctl --more --json=short call /run/systemd/report/io.systemd.CGroup io.systemd.Metrics.List {})
echo "$cgroup_metrics" | grep '"name":"io.systemd.CGroup.CpuUsage"' | grep '"type":"total"' >/dev/null
echo "$cgroup_metrics" | grep '"name":"io.systemd.CGroup.CpuUsage"' | grep '"type":"user"' >/dev/null
echo "$cgroup_metrics" | grep '"name":"io.systemd.CGroup.CpuUsage"' | grep '"type":"system"' >/dev/null
# test io.systemd.Network Metrics
varlinkctl info /run/systemd/report/io.systemd.Network
varlinkctl list-methods /run/systemd/report/io.systemd.Network