diff --git a/README.md b/README.md index 1daecfaae3..bc598012f4 100644 --- a/README.md +++ b/README.md @@ -447,10 +447,25 @@ Cluster performance and capacity can be scaled up in two ways: General recommendations for cluster scalability: -- Adding more CPU and RAM to existing `vmselect` nodes improves the performance for heavy queries, which process big number of time series with big number of raw samples. See [this article on how to detect and optimize heavy queries](https://valyala.medium.com/how-to-optimize-promql-and-metricsql-queries-85a1b75bf986). -- Adding more `vmstorage` nodes increases the number of [active time series](https://docs.victoriametrics.com/FAQ.html#what-is-an-active-time-series) the cluster can handle. This also increases query performance over time series with [high churn rate](https://docs.victoriametrics.com/FAQ.html#what-is-high-churn-rate). The cluster stability is also improved with the number of `vmstorage` nodes, since active `vmstorage` nodes need to handle lower additional workload when some of `vmstorage` nodes become unavailable. -- Adding more CPU and RAM to existing `vmstorage` nodes increases the number of [active time series](https://docs.victoriametrics.com/FAQ.html#what-is-an-active-time-series) the cluster can handle. It is preferred to add more `vmstorage` nodes over adding more CPU and RAM to existing `vmstorage` nodes, since higher number of `vmstorage` nodes increases cluster stability and improves query performance over time series with [high churn rate](https://docs.victoriametrics.com/FAQ.html#what-is-high-churn-rate). +- Adding more CPU and RAM to existing `vmselect` nodes improves the performance for heavy queries, which process big number of time series with big number of raw samples. + See [this article on how to detect and optimize heavy queries](https://valyala.medium.com/how-to-optimize-promql-and-metricsql-queries-85a1b75bf986). + +- Adding more `vmstorage` nodes (aka horizontal scaling) increases the number of [active time series](https://docs.victoriametrics.com/FAQ.html#what-is-an-active-time-series) + the cluster can handle. This also increases query performance over time series with [high churn rate](https://docs.victoriametrics.com/FAQ.html#what-is-high-churn-rate), + since every `vmstorage` node contains lower number of time series when the number of `vmstorage` nodes increases. + + The cluster stability is also improved with the number of `vmstorage` nodes, since active `vmstorage` nodes need to handle lower additional workload + when some of `vmstorage` nodes become unavailable. For example, if one node out of 3 nodes is unavailable, then `1/3=33%` of the load is re-distributed across 2 remaining nodes, + so per-node workload increase is `(1/3/2)/(1/3) = 1/2 = 50%`. + If one node out of 10 nodes is unavailable, then `1/10=10%` of the load is re-distributed across 9 remaining nodes, so per-node workload increase is `(1/10/9)/(1/10) = 1/9 =~ 11%`. + +- Adding more CPU and RAM to existing `vmstorage` nodes (aka vertical scaling) increases the number + of [active time series](https://docs.victoriametrics.com/FAQ.html#what-is-an-active-time-series) the cluster can handle. + It is preferred to add more `vmstorage` nodes over adding more CPU and RAM to existing `vmstorage` nodes, since higher number of `vmstorage` nodes + increases cluster stability and improves query performance over time series with [high churn rate](https://docs.victoriametrics.com/FAQ.html#what-is-high-churn-rate). + - Adding more `vminsert` nodes increases the maximum possible data ingestion speed, since the ingested data may be split among bigger number of `vminsert` nodes. + - Adding more `vmselect` nodes increases the maximum possible queries rate, since the incoming concurrent requests may be split among bigger number of `vmselect` nodes. Steps to add `vmstorage` node: @@ -459,6 +474,8 @@ Steps to add `vmstorage` node: 1. Gradually restart all the `vmselect` nodes with new `-storageNode` arg containing `<new_vmstorage_host>`. 1. Gradually restart all the `vminsert` nodes with new `-storageNode` arg containing `<new_vmstorage_host>`. +In order to handle uneven disk space usage distribution after adding new `vmstorage` node it is possible to update `vminsert` configuration to route newly ingested metrics only to new storage nodes. Once disk usage will be similar configuration can be updated to include all nodes again. Note that `vmselect` nodes need to reference all storage nodes for querying. + ## Updating / reconfiguring cluster nodes All the node types - `vminsert`, `vmselect` and `vmstorage` - may be updated via graceful shutdown.