HPC/nvidia_smi_exporter
8
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Added deviceCount

Browse Source 
Removed Unsupported messages as prometheus fails on that.
This commit is contained in:
Egon Rijpkema 2019-03-12 15:45:26 +01:00
parent 9e786b3456
commit 9c3d7cfd77
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
nvidia_smi_exporter.go
View File

@ -8,6 +8,7 @@ import (
"net/http" "net/http"
"os" "os"
"os/exec" "os/exec"
"strconv"
"strings" "strings"
) )
@ -39,13 +40,21 @@ func metrics(response http.ResponseWriter, request *http.Request) {
"utilization.memory", "memory.total", "memory.free", "memory.used", "eccerrorsaggregate.total", "eccuncorrectedvolatile.total"} "utilization.memory", "memory.total", "memory.free", "memory.used", "eccerrorsaggregate.total", "eccuncorrectedvolatile.total"}
result := "" result := ""
max_id := ""
for _, row := range records { for _, row := range records {
name := fmt.Sprintf("%s[%s]", row[0], row[1]) name := fmt.Sprintf("%s[%s]", row[0], row[1])
for idx, value := range row[2:] { for idx, value := range row[2:] {
result = fmt.Sprintf( result = fmt.Sprintf(
"%s%s{gpu=\"%s\"} %s\n", result, "%s%s{gpu=\"%s\"} %s\n", result,
metricList[idx], name, value) metricList[idx], name, value)
max_id = row[1]
} }
deviceCount := 0
deviceCount, err = strconv.Atoi(max_id)
result = fmt.Sprintf(
"%sdeviceCount %d\n", result, deviceCount+1)
result = strings.Replace(result, "[Not Supported]", "-1", -1)
} }
fmt.Fprintf(response, strings.Replace(result, ".", "_", -1)) fmt.Fprintf(response, strings.Replace(result, ".", "_", -1))