This is an internal documentation. There is a good chance you’re looking for something else. See Disclaimer.

OpenShift Metrics¶

Cluster-Wide Graphs¶

Hint

Only available to members of group tocco-admin.

Hint

SI units are used, thus, 1 GB = 1,000,000,000 bytes = ~0.93 GiB.

To debug a single application, see also:

All pods sorted by memory usage:

oc adm top pod --all-namespace --sort-by cpu

Per node usage:

oc adm top node

To debug a single application, see also:

All pods sorted by memory usage:

oc adm top pod --all-namespace --sort-by memory

Per node usage:

oc adm top node

Link	Short Description	Full Description
graph 1	Total memory	Memory of all worker nodes (=nodes that run apps) Metrics: Available memory Available memory with redundancy [1] Requested memory [2] Used memory
graph 2	Used vs requested	Fraction of requested memory used. Expected to be between 0.8 and 0.9 generally. May be lower after maintenance as Nice requires less memory after a restart. Metrics: maximum 0.95 quantile (=95th percentile) median 0.05 quantile minimum
graph 3	Used vs requested per version	Same as used vs requested above but per Nice minor version. Note that Nice <2.27 used G1GC as garbage collector which did not uncommit (=return to OS) memory.
graph 4	Used vs requested >= v2.26	Same as used vs requested but only for Nice >= v2.26. v2.26 and later use ShanandoahGC.
graph 5	Used vs requested < v2.26	Same as used vs requested but only for Nice < v2.26. Before v2.26 G1GC was used.
graph 6	Used vs requested production	Same as used vs requested but only for production systems.
graph 7	Used vs requested test	Same as used vs requested but only for test systems.
graph 10	Used vs request per installation	Same as used vs requested but per installation.
graph 8	Unused per node	Memory not in active use (but possibly requested)
graph 9	Available per node	Available memory per node. This is, memory that has not been requested [2] and may be used for newly created pods.

Footnotes