Create KB article (Best Practices for Optimizing Longhorn Disk Performance) #51
Conversation
✅ Deploy Preview for harvester-home-preview ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
|
|
||
| For data-intensive applications, you can use pod scheduling functions such as node selector or taint toleration. These functions allow you to schedule the workload to a specific storage-tagged node together with one replica. | ||
|
|
||
| - **Revision counter**: You can disable the [revision counter](https://longhorn.io/docs/1.5.3/advanced-resources/deploy/revision_counter/) to improve IO performance, especially for write-intensive applications. When the revision counter is disabled, Longhorn does not track write operations for replicas and the Longhorn Engine does not check the `revision.counter` file after restarting. |
There was a problem hiding this comment.
@innobead Do we want to recommend this for production environments?
There was a problem hiding this comment.
@innobead Can you comment on this?
There was a problem hiding this comment.
Let's remove this from this practice, as the focus here should be on actions that are 'recommended' be taken.
However, this depends on a very strict requirement already mentioned in the 'important information' below, and it could be challenging to achieve in a prod setup. Additionally, it's not a recommendation but rather an optional method to enhance performance.
There was a problem hiding this comment.
I have removed this item.
|
|
||
| - **Recurring snapshots**: Periodically clean up system-generated snapshots and retain only the number of snapshots that makes sense for your implementation. For applications with replication capability, periodically [delete all types of snapshots](https://longhorn.io/docs/1.5.3/concepts/#243-deleting-snapshots). | ||
|
|
||
| - **Recurring filesystem trim**: Periodically [trim the filesystem](https://longhorn.io/docs/1.5.3/volumes-and-nodes/trim-filesystem/) inside volumes to reclaim disk space. |
There was a problem hiding this comment.
@innobead I suggest we remove this. Most harvester VMs use block mode volume, I'm not sure if the trim operation works here or not. (Probably doing a trim inside VM?)
There was a problem hiding this comment.
@innobead Can you comment on this?
There was a problem hiding this comment.
Agreed, let's remove this as it only works for the fs mount point (volume mode: fs).
Longhorn volume supports unmap, which should continue functioning if users perform trim on the VM fs. However, this should not be recommended, as there is a known issue at harvester/harvester#4739. Additionally, the user experience is not good.
There was a problem hiding this comment.
I have removed this item.
There was a problem hiding this comment.
In general, LGTM. After resolve the comments, we can merge this.
However, we continue improving this doc over time, since there will be some new practices to be introduced later. For example, snapshot space management, default priority class, etc.
@jillian-maroket We should have one in Longhorn doc as well, but it should be in a dedicated in https://longhorn.io/docs/1.6.0/best-practices/. Please help create a ticket for that and plan it.
This KB article addresses a customer request for best practice recommendations regarding Harvester configuration with disk performance in mind.
Jira: SURE-6729
GitHub issue: harvester/harvester#3356
David Ko and the Longhorn engineering team provided most of the content. I added supporting details from the Longhorn documentation.