Xiaochao Dong
Xiaochao Dong
@jesusvazquez I have run the [benchmark test](https://github.com/prometheus/prometheus/blob/4b9f248e8569fc5163841c2e21b8fa5d3a61b311/tsdb/head_test.go#L134-L263) against the `main` branch, and interestingly there is no significant difference on my old mac. I may look into this later.
The benchmark result looks better on my mac if I modify the test like this: ```patch @@ -185,9 +185,13 @@ func BenchmarkLoadWAL(b *testing.B) { // Write series. refSeries := make([]record.RefSeries,...
> "we need a little more memory to get the task done" How much are we talking? @richardkiene I think there is no significant difference in memory consumption. I'd expect...
I have done some tests manually with more than 1 million series (7 checkpoint segments, 27 WAL segments). The steps are as follows: 1. The workload definition: https://gist.github.com/damnever/a4648665737b2d0366cce6a9b255d1bc 2. Deploy...
Failed tests seem to be unrelated.
Friendly ping @codesome, is there anything we should do here?
Thanks, @colega. I would like to hear some advice from @codesome before I commit more changes.. 😅
We could get invalid labels in multiple ways, I will describe a common one here: - use prometheus A to scrape the metrics from Kubernetes resources, relabel `__meta_kubernetes_X_label_` with `labelmap`,...
@roidelapluie > Relabeling can drop labels. I think this is a subtly breaking change: > > 1. You could have invalid labels and drop them before they even touch the...
Tornado4.4 removed support for Python 2.6.