Details
-
Task
-
Resolution: Incomplete
-
Major
-
2.0.2
-
None
-
Red Hat OpenSHift
-
1
Description
It seems there are missing pre requisites on OCP 4.x inside the documentation.
While building a fresh new OpenShift OVCP 4.6 cluster, no specific pre requisites are mentioned inside Couchbase documentation and the cluster did not work well (rebalance failure, impossible to perform a cbcollect_info).
Some pre requisites like ulimit definition (nproc, nofile), fs.file_max apparently needs to be defined at Openshift cluster level (not Couchbase Server pods level, i.e. not in CouchbaseCluster.yaml file).
Problems:
1. for nproc, what value put-something-big-here should we set in:
limits:
nproc:
hard: put-something-big-here
soft: put-something-big-here
2. Are there any other pre requisites about OCP4 to apply for Couchbase Server to work properly?
3. In case the Sync Gateway is deployed on OCP 4, are there other specific pre requisites to apply too ?
There are OS tuning for SGW deployment on bare-metal/VM machines. See https://docs.couchbase.com/sync-gateway/current/os-level-tuning.html#linux-instructions-centos
=> Are there anything else to be done at SGW YAML file (or again at Openshift cluster level to make it work properly) ?
For example, there is no mention of fs.file_max nor ulimit settings (used for setting file descriptor number this time) in the[ Deploying a Sync Gateway Cluster|https://docs.couchbase.com/sync-gateway/2.7/kubernetes/deploy-cluster.html] or Connecting Sync Gateway to a Couchbase Cluster documentation pages.
The only setting that could be changed (in SGW YAML file) would be maxFileDescriptors but it is useless to set it if fs.file-max is not increased too, right?
Read also : https://access.redhat.com/solutions/5366631
Attachments
Issue Links
- relates to
-
K8S-2091 Log collection issues with pod/exec and in-memory clusters
- Closed