Tips for troubleshooting issues. First of all check your resources. Make sure you are looking at the correct resources. Remember to specify the -n for namespace and you can also get a list of all resources from a given namespace.
Some common problems are:
Lack of resources in the k8s cluster
Wrong node affinity, taints and tollerations
Wrong regions / availability zones
Some common performance issues are:
Absence of cpuset, hostNetworking, wrong CPU manager policy in kubelet, non-guaranteed QoS class (limits != requests)
Absence of monitoring
Bad practices on the client-side. Check out the ScyllaDB University lesson on Data Modeling and How to Write Better Apps.
Transcript
A few words about troubleshooting, also always make sure you’re looking at the correct resources so
always use namespaces for resources, I’m going full screen, for your resources you can
use ‘get all’ to get basic view of the entire deployment, if something isn’t working
make sure to check out the events, inside the namespace and
you may also check the events generated by the ScyllaDB cluster itself
or if you don’t find anything there, you can also check the logs
of the pods and logs from the ScyllaDB Operator, so maybe there will be something there
if you don’t have enough performance make sure you set the cpuset and hostNetworking
also make sure to set up the static CPU manager in kubelet and make sure that your resources, limits
and requests are equal, so these are the quick tips that help with performance and also always
use the monitoring, because most of the issues from the application side can be seen there
and for other stuff you can also check the training session
‘How to Write Better Apps’ on our ScyllaDB University
