prometheus/cadvisor
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Add support for identifying OOM victim.

Browse Source
This commit is contained in:
Victor Marmol 2015-04-15 13:09:22 -07:00
parent a59bfbbd5c
commit 32b33b215a
2 changed files with 21 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
utils/oomparser/oomparser.go
View File

@ -30,7 +30,7 @@ import (
)
var containerRegexp *regexp.Regexp = regexp.MustCompile(
`Task in (.*) killed as a result of limit of `)
`Task in (.*) killed as a result of limit of (.*)`)
var lastLineRegexp *regexp.Regexp = regexp.MustCompile(
`(^[A-Z]{1}[a-z]{2} .*[0-9]{1,2} [0-9]{1,2}:[0-9]{2}:[0-9]{2}) .* Killed process ([0-9]+) \(([0-9A-Za-z_]+)\)`)
var firstLineRegexp *regexp.Regexp = regexp.MustCompile(
@ -52,6 +52,9 @@ type OomInstance struct {
TimeOfDeath time.Time
// the absolute name of the container that OOMed
ContainerName string
// the absolute name of the container that was killed
// due to the OOM.
VictimContainerName string
}
// gets the container name from a line and adds it to the oomInstance.
@ -61,6 +64,7 @@ func getContainerName(line string, currentOomInstance *OomInstance) error {
return nil
}
currentOomInstance.ContainerName = path.Join("/", parsedLine[1])
currentOomInstance.VictimContainerName = path.Join("/", parsedLine[2])
return nil
}

10
utils/oomparser/oomparser_test.go
View File

@ -17,13 +17,14 @@ package oomparser
import (
"bufio"
"os"
"reflect"
"testing"
"time"
)
const startLine = "Jan 21 22:01:49 localhost kernel: [62278.816267] ruby invoked oom-killer: gfp_mask=0x201da, order=0, oom_score_adj=0"
const endLine = "Jan 21 22:01:49 localhost kernel: [62279.421192] Killed process 19667 (evilprogram2) total-vm:1460016kB, anon-rss:1414008kB, file-rss:4kB"
const containerLine = "Jan 26 14:10:07 kateknister0.mtv.corp.google.com kernel: [1814368.465205] Task in /mem2 killed as a result of limit of /mem2"
const containerLine = "Jan 26 14:10:07 kateknister0.mtv.corp.google.com kernel: [1814368.465205] Task in /mem2 killed as a result of limit of /mem3"
const containerLogFile = "containerOomExampleLog.txt"
const systemLogFile = "systemOomExampleLog.txt"
@ -39,6 +40,7 @@ func createExpectedContainerOomInstance(t *testing.T) *OomInstance {
ProcessName: "memorymonster",
TimeOfDeath: deathTime,
ContainerName: "/mem2",
VictimContainerName: "/mem3",
}
}
@ -54,6 +56,7 @@ func createExpectedSystemOomInstance(t *testing.T) *OomInstance {
ProcessName: "badsysprogram",
TimeOfDeath: deathTime,
ContainerName: "/",
VictimContainerName: "/",
}
}
@ -73,6 +76,9 @@ func TestGetContainerName(t *testing.T) {
if currentOomInstance.ContainerName != "/mem2" {
t.Errorf("getContainerName should have set containerName to /mem2, not %s", currentOomInstance.ContainerName)
}
if currentOomInstance.VictimContainerName != "/mem3" {
t.Errorf("getContainerName should have set victimContainerName to /mem3, not %s", currentOomInstance.VictimContainerName)
}
}
func TestGetProcessNamePid(t *testing.T) {
@ -139,7 +145,7 @@ func helpTestStreamOoms(oomCheckInstance *OomInstance, sysFile string, t *testin
select {
case oomInstance := <-outStream:
if *oomCheckInstance != *oomInstance {
if reflect.DeepEqual(*oomCheckInstance, *oomInstance) {
t.Errorf("wrong instance returned. Expected %v and got %v",
oomCheckInstance, oomInstance)
}