mirror of
https://github.com/VictoriaMetrics/VictoriaMetrics.git
synced 2024-12-15 08:23:34 +01:00
1a254ea20c
This should improve vmselect performance scalability on systems with many CPU cores.
The following tasks were done:
- Use separate temporary files for storing the data read from each vmstorage node.
This may result in the following potential issues:
- Up to N times higher memory usage for performing each query where N is the number
of vmstorage nodes known to vmselect.
This issue shouldn't increase chances of out of memory errors in most cases,
since per-query memory overhead is quite low comparing to the overall vmselect memory usage.
- Up to N times higher number of open temporary files where N is the number
of vmstorage nodes known to vmselect.
This issue should be fixed by increasing the limit on the number of open files.
- Use separate counters per each vmstorage node for various stats calculation
when reading the data from vmstorage nodes.
|
||
|---|---|---|
| .. | ||
| bufferedwriter | ||
| clusternative | ||
| deployment | ||
| graphite | ||
| graphiteql | ||
| multiarch | ||
| netstorage | ||
| prometheus | ||
| promql | ||
| querystats | ||
| searchutils | ||
| vmui | ||
| main.go | ||
| Makefile | ||
| README.md | ||
vmselect performs the following tasks:
-
Splits incoming selects to tasks for
vmstoragenodes and issues these tasks to all thevmstoragenodes in the cluster. -
Merges responses from all the
vmstoragenodes and returns a single response.
The vmui directory contains static contents built from app/vmui package with make vmui-update command. The vmui page is available at http://<vmselect>:8481/select/<accountID>/vmui/.