add user guide #187
add user guide #187
Conversation
|
This looks great! I will review it some more, but just wanted to add a note on structure. I would like to have an operations guide that assumes everything is installed and working. Its odd for someone using GKE to see a section on local volume provisioner. Can we split this user guide to installation and operation? |
|
Users on GKE also need to deploy local volume provisioner if they want to use local volume, but the setup is a bit different. I'll split this document into two files. |
docs/operation-guide.md
Outdated
|
|
||
| ## Restore | ||
|
|
||
| Currently, tidb-operator only supports restoring from full backup in GCS bucket. The `restore` section in charts/tidb-cluster/values.yaml should have enough comments as document. |
There was a problem hiding this comment.
I don't think the restore workflow makes sense. The helm chart is for persistent objects, but restore is a one-off operation. There could be multiple restores of different databases or different restores at different times, and I don't know how helm upgrade will handle it. Can we separate out the restore operation from the helm chart?
There was a problem hiding this comment.
The restore section itself documents what fields can be changed, but not any of the workflow concerns.
|
LGTM. I have some comments, but I would also be happy to merge and address them afterwards. |
|
|
||
| ## Scale TiDB cluster | ||
|
|
||
| TiDB Operator has full support of horizontal scaling. But for vertical scaling, if you're using local volumes for PD and TiKV, then scaling up may cause pod pending if the node doesn't have enough resources. So it's not recommended to do vertical scaling. |
There was a problem hiding this comment.
What about vertical scale-down?
There was a problem hiding this comment.
Though scaling down is doable. Vertical scaling is not recommended, users have to adjust some configurations for TiDB when scaling vertically to make it have better performance. Besides scaling down will kill the pod one by one other than in-place scale down which might not what user expected.
There was a problem hiding this comment.
What about TiDB SQL? I should be able to scale those vertically?
I created a github issue focused on TiKV/PD here: #191
There was a problem hiding this comment.
I've added the recommendations for TiDB cluster deployment.
There was a problem hiding this comment.
A lot more thorough now, thanks! I still don't know from the guide if I can safely vertical scale TIDB SQL.
|
I think we should merge this PR and fix other issues in following PRs. |
|
Yes, good to merge |
* add user guide * split user guide into tidb-operator setup, tidb-cluster operation and troubleshooting * add deployment recommendation for TiDB cluster * update README for document links
This PR adds a general user guide for tidb-operator. It closes #161