Hide Forgot
Document URL: https://access.redhat.com/documentation/en/openshift-container-platform/3.3/single/installation-and-configuration/#multi-masters-using-native-ha Section Number and Name: 2.1.4.3. Multiple Masters Using Native HA Describe the issue: It does not become clear from that section that there is an etcd that is embedded into the openshift binary and that it is this etcd that can not be clustered. Readers might be mislead to believe that you need to run additional infrastructure nodes to carry the etcd in the HA environment. Suggestions for improvement: - Explain clearly that there is an etcd embedded in the openshift binary - HA clustering of Etcd requires the running etcd in separate processes - these are best run on the master nodes. Additional information:
Lutz, There seems to be an issue with this section specifically, where some updates I made recently are not there. You can see them with this PR: https://github.com/openshift/openshift-docs/pull/2826/files There is a note box that's underneath the hosts table that sounds suspiciously like what you might be asking for... However, I'm not too sure if this fulfills this BZ. I do want to ask if what's in the section introduction for the section fulfills this BZ, and was maybe looked over when you reported this: https://access.redhat.com/documentation/en/openshift-container-platform/3.3/single/installation-and-configuration/#multiple-masters It's my understanding that it doesn't matter too much if the etcd is embedded or external, as long as it's signaled in the ansible file. I'm not actually sure what you mean by "it is this etcd that can not be clustered". The default OpenShift etcd should not be used in which way exactly? It's also probably worth mentioning that these are guidelines and examples, and not necessarily every example under the sun. That might be a little to difficult to document. I hope I'm not misunderstanding the request. If there's any clarifications you could send my way, that'd be great.
Update: The PR seems to be live now.
Hi brice, the best practice that I have been told how to run HA is to have the etcd on the masters as a separate process. Section 2.1.4.3. Multiple Masters Using Native HA, lists etcd as separate notes. Cloud you include a szenario with multiple master running etcd as well? Also for etcd there this the option of using an embedded etcd in the openshift binary. This etcd can not be clustered. It did not become clear what the embedded etcd is really. Regards Lutz
Lutz, After talking to Jason in the install team, I've submitted a PR on just what they recommend: https://github.com/openshift/openshift-docs/pull/3277 Can I ask that this fulfills this BZ? I feel like embedded etcd is not something that is recommended, so I've taken away some of the words asking for that.
Commit pushed to master at https://github.com/openshift/openshift-docs https://github.com/openshift/openshift-docs/commit/bbe271ed1b08c2e1e06ce196e2c88b905ff30a3b Merge pull request #3277 from bfallonf/hosts-1384994 Bug 1384994 edited information on embedded etcd
Link to released docs: https://access.redhat.com/documentation/en/openshift-container-platform/3.4/single/installation-and-configuration/#downgrade-verifying-backups https://access.redhat.com/documentation/en/openshift-container-platform/3.4/single/installation-and-configuration/#single-master https://access.redhat.com/documentation/en/openshift-container-platform/3.4/single/installation-and-configuration/#environment-scenarios