Arnike/watcher
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

vm workload consolidation: use actual host metrics

Browse Source 
The "vm workload consolidation" strategy is summing up instance
usage in order to estimate host usage.

The problem is that some infrastructure services (e.g. OVS or Ceph
clients) may also use a significant amount of resources, which
would be ignored. This can impact Watcher's ability to detect
overloaded nodes and correctly rebalance the workload.

This commit will use the host metrics, if available. The proposed
implementation uses the maximum value between the host metric
and the sum of the instance metrics.

Note that we're holding a dict of host metric deltas in order to
account for planned migrations.

Change-Id: I82f474ee613f6c9a7c0a9d24a05cba41d2f68edb
This commit is contained in:
Lucian Petrut
2023-10-17 16:09:19 +03:00
parent 40e93407c7
commit 424e9a76af
4 changed files with 160 additions and 96 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
doc/source/strategies/vm_workload_consolidation.rst
View File

@@ -26,9 +26,15 @@ metric service name plugins comment
``memory.resident`` ceilometer_ none
``memory`` ceilometer_ none
``disk.root.size`` ceilometer_ none
``compute.node.cpu.percent`` ceilometer_ none (optional) need to set the
``compute_monitors`` option
to ``cpu.virt_driver`` in the
nova.conf.
``hardware.memory.used`` ceilometer_ SNMP_ (optional)
============================ ============ ======= =========================
.. _ceilometer: https://docs.openstack.org/ceilometer/latest/admin/telemetry-measurements.html#openstack-compute
.. _SNMP: https://docs.openstack.org/ceilometer/latest/admin/telemetry-measurements.html#snmp-based-meters
Cluster data model
******************