StackLight known issues¶
This section lists the StackLight known issues with workarounds for the Mirantis OpenStack for Kubernetes release 21.2.
[13233] Low memory limits for StackLight Helm Controller cause update failure
[12917] prometheus-tf-vrouter-exporter pods fail to start on TF nodes with DPDK
[13233] Low memory limits for StackLight Helm Controller cause update failure¶
During the MOS cluster update to Cluster release
6.14.0, StackLight Helm Controller containers (controller
and/or
tiller
) may get OOMKilled and cause failure to update.
As a workaround, manually increase the default resource requests and limits
for stacklightHelmControllerController
and
stacklightHelmControllerTiller
in the StackLight Helm chart values of the
Cluster release resource:
resources:
stacklightHelmControllerController:
requests:
cpu: "250m"
memory: "128Mi"
limits:
cpu: "1000m"
memory: "512Mi"
stacklightHelmControllerTiller:
requests:
cpu: "250m"
memory: "128Mi"
limits:
cpu: "1000m"
memory: "2048Mi"
For details about resource limits, see Mirantis Container Cloud Operations Guide: Resource limits.
[12917] prometheus-tf-vrouter-exporter pods fail to start on TF nodes with DPDK¶
StackLight deploys the prometheus-tf-vrouter-exporter
exporter based on the
node selector matching the tfvrouter: enabled
node label. The Tungsten
Fabric nodes with DPDK have the tfvrouter-dpdk: enabled
label set instead.
Therefore, the prometheus-tf-vrouter-exporter
exporter fails to start on
these nodes.
Workaround:
Add the
tfvrouter-fix: enabled
label to every node that contains either thetfvrouter: enabled
or thetfvrouter-dpdk: enabled
node label.kubectl label node <node_name> tfvrouter-fix=enabled
In the Cluster release resource, specify the following
nodeSelector
definition in the StackLight Helm chart values:nodeSelector: component: tfVrouterExporter: tfvrouter-fix: enabled
Once done, prometheus-tf-vrouter-exporter
will be deployed to every node
with the tfvrouter-fix: enabled
label.