2016-04-19 01:10:00 -04:00
|
|
|
// +build linux
|
|
|
|
|
|
|
|
|
|
/*
|
2016-06-02 20:25:58 -04:00
|
|
|
Copyright 2016 The Kubernetes Authors.
|
2016-04-19 01:10:00 -04:00
|
|
|
|
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
|
you may not use this file except in compliance with the License.
|
|
|
|
|
You may obtain a copy of the License at
|
|
|
|
|
|
|
|
|
|
http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
|
|
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
|
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
|
See the License for the specific language governing permissions and
|
|
|
|
|
limitations under the License.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
package util
|
|
|
|
|
|
|
|
|
|
import (
|
|
|
|
|
"errors"
|
|
|
|
|
"os"
|
2017-12-13 11:31:38 -05:00
|
|
|
"reflect"
|
Avoid deleted iSCSI LUNs in the kernel
This change ensures that iSCSI block devices are deleted after
unmounting, and implements scanning of individual LUNs rather
than scanning the whole iSCSI bus.
In cases where an iSCSI bus is in use by more than one attachment,
detaching used to leave behind phantom block devices, which could
cause I/O errors, long timeouts, or even corruption in the case
when the underlying LUN number was recycled. This change makes
sure to flush references to the block devices after unmounting.
The original iSCSI code scanned the whole target every time a LUN
was attached. On storage controllers that export multiple LUNs on
the same target IQN, this led to a situation where nodes would
see SCSI disks that they weren't supposed to -- possibly dozens or
hundreds of extra SCSI disks. This caused 3 significant problems:
1) The large number of disks wasted resources on the node and
caused a minor drag on performance.
2) The scanning of all the devices caused a huge number of uevents
from the kernel, causing udev to bog down for multiple minutes in
some cases, triggering timeouts and other transient failures.
3) Because Kubernetes was not tracking all the "extra" LUNs that
got discovered, they would not get cleaned up until the last LUN
on a particular target was detached, causing a logout. This led
to significant complications:
In the time window between when a LUN was unintentially scanned,
and when it was removed due to a logout, if it was deleted on the
backend, a phantom reference remained on the node. In the best
case, the phantom LUN would cause I/O errors and timeouts in the
udev system. In the worst case, the backend could reuse the LUN
number for a new volume, and if that new volume were to be
scheduled to a pod with a phantom reference to the old LUN by the
same number, the initiator could get confused and possibly corrupt
data on that volume.
To avoid these problems, the new implementation only scans for
the specific LUN number it expects to see. It's worth noting that
the default behavior of iscsiadm is to automatically scan the
whole bus on login. That behavior can be disabled by setting
node.session.scan = manual
in iscsid.conf, and for the reasons mentioned above, it is
strongly recommended to set that option. This change still works
regardless of the setting in iscsid.conf, and while automatic
scanning will cause some problems, this change doesn't make the
problems any worse, and can make things better in some cases.
2018-07-24 23:58:19 -04:00
|
|
|
"regexp"
|
2016-04-19 01:10:00 -04:00
|
|
|
"testing"
|
|
|
|
|
"time"
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
type mockOsIOHandler struct{}
|
|
|
|
|
|
Avoid deleted iSCSI LUNs in the kernel
This change ensures that iSCSI block devices are deleted after
unmounting, and implements scanning of individual LUNs rather
than scanning the whole iSCSI bus.
In cases where an iSCSI bus is in use by more than one attachment,
detaching used to leave behind phantom block devices, which could
cause I/O errors, long timeouts, or even corruption in the case
when the underlying LUN number was recycled. This change makes
sure to flush references to the block devices after unmounting.
The original iSCSI code scanned the whole target every time a LUN
was attached. On storage controllers that export multiple LUNs on
the same target IQN, this led to a situation where nodes would
see SCSI disks that they weren't supposed to -- possibly dozens or
hundreds of extra SCSI disks. This caused 3 significant problems:
1) The large number of disks wasted resources on the node and
caused a minor drag on performance.
2) The scanning of all the devices caused a huge number of uevents
from the kernel, causing udev to bog down for multiple minutes in
some cases, triggering timeouts and other transient failures.
3) Because Kubernetes was not tracking all the "extra" LUNs that
got discovered, they would not get cleaned up until the last LUN
on a particular target was detached, causing a logout. This led
to significant complications:
In the time window between when a LUN was unintentially scanned,
and when it was removed due to a logout, if it was deleted on the
backend, a phantom reference remained on the node. In the best
case, the phantom LUN would cause I/O errors and timeouts in the
udev system. In the worst case, the backend could reuse the LUN
number for a new volume, and if that new volume were to be
scheduled to a pod with a phantom reference to the old LUN by the
same number, the initiator could get confused and possibly corrupt
data on that volume.
To avoid these problems, the new implementation only scans for
the specific LUN number it expects to see. It's worth noting that
the default behavior of iscsiadm is to automatically scan the
whole bus on login. That behavior can be disabled by setting
node.session.scan = manual
in iscsid.conf, and for the reasons mentioned above, it is
strongly recommended to set that option. This change still works
regardless of the setting in iscsid.conf, and while automatic
scanning will cause some problems, this change doesn't make the
problems any worse, and can make things better in some cases.
2018-07-24 23:58:19 -04:00
|
|
|
func (handler *mockOsIOHandler) ReadFile(filename string) ([]byte, error) {
|
|
|
|
|
portPattern := regexp.MustCompile("^/sys/class/iscsi_host/(host\\d)/device/session\\d/connection\\d:0/iscsi_connection/connection\\d:0/(?:persistent_)?port$")
|
|
|
|
|
if portPattern.MatchString(filename) {
|
|
|
|
|
return []byte("3260"), nil
|
|
|
|
|
}
|
|
|
|
|
addressPattern := regexp.MustCompile("^/sys/class/iscsi_host/(host\\d)/device/session\\d/connection\\d:0/iscsi_connection/connection\\d:0/(?:persistent_)?address$")
|
|
|
|
|
matches := addressPattern.FindStringSubmatch(filename)
|
|
|
|
|
if nil != matches {
|
|
|
|
|
switch matches[1] {
|
|
|
|
|
case "host2":
|
|
|
|
|
return []byte("10.0.0.1"), nil
|
|
|
|
|
case "host3":
|
|
|
|
|
return []byte("10.0.0.2"), nil
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
targetNamePattern := regexp.MustCompile("^/sys/class/iscsi_host/(host\\d)/device/session\\d/iscsi_session/session\\d/targetname$")
|
|
|
|
|
matches = targetNamePattern.FindStringSubmatch(filename)
|
|
|
|
|
if nil != matches {
|
|
|
|
|
switch matches[1] {
|
|
|
|
|
case "host2":
|
|
|
|
|
return []byte("target1"), nil
|
|
|
|
|
case "host3":
|
|
|
|
|
return []byte("target2"), nil
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return nil, errors.New("Not Implemented for Mock")
|
|
|
|
|
}
|
|
|
|
|
|
2016-04-19 01:10:00 -04:00
|
|
|
func (handler *mockOsIOHandler) ReadDir(dirname string) ([]os.FileInfo, error) {
|
|
|
|
|
switch dirname {
|
2017-12-13 11:31:38 -05:00
|
|
|
case "/sys/block/dm-1/slaves":
|
|
|
|
|
f1 := &fakeFileInfo{
|
2016-04-19 01:10:00 -04:00
|
|
|
name: "sda",
|
|
|
|
|
}
|
2017-12-13 11:31:38 -05:00
|
|
|
f2 := &fakeFileInfo{
|
|
|
|
|
name: "sdb",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1, f2}, nil
|
2016-04-19 01:10:00 -04:00
|
|
|
case "/sys/block/":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "sda",
|
|
|
|
|
}
|
|
|
|
|
f2 := &fakeFileInfo{
|
|
|
|
|
name: "dm-1",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1, f2}, nil
|
Avoid deleted iSCSI LUNs in the kernel
This change ensures that iSCSI block devices are deleted after
unmounting, and implements scanning of individual LUNs rather
than scanning the whole iSCSI bus.
In cases where an iSCSI bus is in use by more than one attachment,
detaching used to leave behind phantom block devices, which could
cause I/O errors, long timeouts, or even corruption in the case
when the underlying LUN number was recycled. This change makes
sure to flush references to the block devices after unmounting.
The original iSCSI code scanned the whole target every time a LUN
was attached. On storage controllers that export multiple LUNs on
the same target IQN, this led to a situation where nodes would
see SCSI disks that they weren't supposed to -- possibly dozens or
hundreds of extra SCSI disks. This caused 3 significant problems:
1) The large number of disks wasted resources on the node and
caused a minor drag on performance.
2) The scanning of all the devices caused a huge number of uevents
from the kernel, causing udev to bog down for multiple minutes in
some cases, triggering timeouts and other transient failures.
3) Because Kubernetes was not tracking all the "extra" LUNs that
got discovered, they would not get cleaned up until the last LUN
on a particular target was detached, causing a logout. This led
to significant complications:
In the time window between when a LUN was unintentially scanned,
and when it was removed due to a logout, if it was deleted on the
backend, a phantom reference remained on the node. In the best
case, the phantom LUN would cause I/O errors and timeouts in the
udev system. In the worst case, the backend could reuse the LUN
number for a new volume, and if that new volume were to be
scheduled to a pod with a phantom reference to the old LUN by the
same number, the initiator could get confused and possibly corrupt
data on that volume.
To avoid these problems, the new implementation only scans for
the specific LUN number it expects to see. It's worth noting that
the default behavior of iscsiadm is to automatically scan the
whole bus on login. That behavior can be disabled by setting
node.session.scan = manual
in iscsid.conf, and for the reasons mentioned above, it is
strongly recommended to set that option. This change still works
regardless of the setting in iscsid.conf, and while automatic
scanning will cause some problems, this change doesn't make the
problems any worse, and can make things better in some cases.
2018-07-24 23:58:19 -04:00
|
|
|
case "/sys/class/iscsi_host":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "host2",
|
|
|
|
|
}
|
|
|
|
|
f2 := &fakeFileInfo{
|
|
|
|
|
name: "host3",
|
|
|
|
|
}
|
|
|
|
|
f3 := &fakeFileInfo{
|
|
|
|
|
name: "ignore",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1, f2, f3}, nil
|
|
|
|
|
case "/sys/class/iscsi_host/host2/device":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "session1",
|
|
|
|
|
}
|
|
|
|
|
f2 := &fakeFileInfo{
|
|
|
|
|
name: "ignore",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1, f2}, nil
|
|
|
|
|
case "/sys/class/iscsi_host/host3/device":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "session2",
|
|
|
|
|
}
|
|
|
|
|
f2 := &fakeFileInfo{
|
|
|
|
|
name: "ignore",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1, f2}, nil
|
|
|
|
|
case "/sys/class/iscsi_host/host2/device/session1":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "connection1:0",
|
|
|
|
|
}
|
|
|
|
|
f2 := &fakeFileInfo{
|
|
|
|
|
name: "ignore",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1, f2}, nil
|
|
|
|
|
case "/sys/class/iscsi_host/host3/device/session2":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "connection2:0",
|
|
|
|
|
}
|
|
|
|
|
f2 := &fakeFileInfo{
|
|
|
|
|
name: "ignore",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1, f2}, nil
|
|
|
|
|
case "/sys/class/iscsi_host/host2/device/session1/target2:0:0/2:0:0:1/block":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "sda",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1}, nil
|
|
|
|
|
case "/sys/class/iscsi_host/host2/device/session1/target2:0:0/2:0:0:2/block":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "sdc",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1}, nil
|
|
|
|
|
case "/sys/class/iscsi_host/host3/device/session2/target3:0:0/3:0:0:1/block":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "sdb",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1}, nil
|
|
|
|
|
case "/sys/class/iscsi_host/host3/device/session2/target3:0:0/3:0:0:2/block":
|
|
|
|
|
f1 := &fakeFileInfo{
|
|
|
|
|
name: "sdd",
|
|
|
|
|
}
|
|
|
|
|
return []os.FileInfo{f1}, nil
|
2016-04-19 01:10:00 -04:00
|
|
|
}
|
Avoid deleted iSCSI LUNs in the kernel
This change ensures that iSCSI block devices are deleted after
unmounting, and implements scanning of individual LUNs rather
than scanning the whole iSCSI bus.
In cases where an iSCSI bus is in use by more than one attachment,
detaching used to leave behind phantom block devices, which could
cause I/O errors, long timeouts, or even corruption in the case
when the underlying LUN number was recycled. This change makes
sure to flush references to the block devices after unmounting.
The original iSCSI code scanned the whole target every time a LUN
was attached. On storage controllers that export multiple LUNs on
the same target IQN, this led to a situation where nodes would
see SCSI disks that they weren't supposed to -- possibly dozens or
hundreds of extra SCSI disks. This caused 3 significant problems:
1) The large number of disks wasted resources on the node and
caused a minor drag on performance.
2) The scanning of all the devices caused a huge number of uevents
from the kernel, causing udev to bog down for multiple minutes in
some cases, triggering timeouts and other transient failures.
3) Because Kubernetes was not tracking all the "extra" LUNs that
got discovered, they would not get cleaned up until the last LUN
on a particular target was detached, causing a logout. This led
to significant complications:
In the time window between when a LUN was unintentially scanned,
and when it was removed due to a logout, if it was deleted on the
backend, a phantom reference remained on the node. In the best
case, the phantom LUN would cause I/O errors and timeouts in the
udev system. In the worst case, the backend could reuse the LUN
number for a new volume, and if that new volume were to be
scheduled to a pod with a phantom reference to the old LUN by the
same number, the initiator could get confused and possibly corrupt
data on that volume.
To avoid these problems, the new implementation only scans for
the specific LUN number it expects to see. It's worth noting that
the default behavior of iscsiadm is to automatically scan the
whole bus on login. That behavior can be disabled by setting
node.session.scan = manual
in iscsid.conf, and for the reasons mentioned above, it is
strongly recommended to set that option. This change still works
regardless of the setting in iscsid.conf, and while automatic
scanning will cause some problems, this change doesn't make the
problems any worse, and can make things better in some cases.
2018-07-24 23:58:19 -04:00
|
|
|
return nil, errors.New("Not Implemented for Mock")
|
2016-04-19 01:10:00 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (handler *mockOsIOHandler) Lstat(name string) (os.FileInfo, error) {
|
|
|
|
|
links := map[string]string{
|
|
|
|
|
"/sys/block/dm-1/slaves/sda": "sda",
|
|
|
|
|
"/dev/sda": "sda",
|
Avoid deleted iSCSI LUNs in the kernel
This change ensures that iSCSI block devices are deleted after
unmounting, and implements scanning of individual LUNs rather
than scanning the whole iSCSI bus.
In cases where an iSCSI bus is in use by more than one attachment,
detaching used to leave behind phantom block devices, which could
cause I/O errors, long timeouts, or even corruption in the case
when the underlying LUN number was recycled. This change makes
sure to flush references to the block devices after unmounting.
The original iSCSI code scanned the whole target every time a LUN
was attached. On storage controllers that export multiple LUNs on
the same target IQN, this led to a situation where nodes would
see SCSI disks that they weren't supposed to -- possibly dozens or
hundreds of extra SCSI disks. This caused 3 significant problems:
1) The large number of disks wasted resources on the node and
caused a minor drag on performance.
2) The scanning of all the devices caused a huge number of uevents
from the kernel, causing udev to bog down for multiple minutes in
some cases, triggering timeouts and other transient failures.
3) Because Kubernetes was not tracking all the "extra" LUNs that
got discovered, they would not get cleaned up until the last LUN
on a particular target was detached, causing a logout. This led
to significant complications:
In the time window between when a LUN was unintentially scanned,
and when it was removed due to a logout, if it was deleted on the
backend, a phantom reference remained on the node. In the best
case, the phantom LUN would cause I/O errors and timeouts in the
udev system. In the worst case, the backend could reuse the LUN
number for a new volume, and if that new volume were to be
scheduled to a pod with a phantom reference to the old LUN by the
same number, the initiator could get confused and possibly corrupt
data on that volume.
To avoid these problems, the new implementation only scans for
the specific LUN number it expects to see. It's worth noting that
the default behavior of iscsiadm is to automatically scan the
whole bus on login. That behavior can be disabled by setting
node.session.scan = manual
in iscsid.conf, and for the reasons mentioned above, it is
strongly recommended to set that option. This change still works
regardless of the setting in iscsid.conf, and while automatic
scanning will cause some problems, this change doesn't make the
problems any worse, and can make things better in some cases.
2018-07-24 23:58:19 -04:00
|
|
|
"/sys/class/iscsi_host/host2/device/session1/target2:0:0/2:0:0:1": "2:0:0:1",
|
|
|
|
|
"/sys/class/iscsi_host/host2/device/session1/target2:0:0/2:0:0:2": "2:0:0:2",
|
|
|
|
|
"/sys/class/iscsi_host/host3/device/session2/target3:0:0/3:0:0:1": "3:0:0:1",
|
|
|
|
|
"/sys/class/iscsi_host/host3/device/session2/target3:0:0/3:0:0:2": "3:0:0:2",
|
2016-04-19 01:10:00 -04:00
|
|
|
}
|
|
|
|
|
if dev, ok := links[name]; ok {
|
|
|
|
|
return &fakeFileInfo{name: dev}, nil
|
|
|
|
|
}
|
|
|
|
|
return nil, errors.New("Not Implemented for Mock")
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (handler *mockOsIOHandler) EvalSymlinks(path string) (string, error) {
|
|
|
|
|
links := map[string]string{
|
2018-10-05 15:59:38 -04:00
|
|
|
"/returns/a/dev": "/dev/sde",
|
|
|
|
|
"/returns/non/dev": "/sys/block",
|
2016-04-19 01:10:00 -04:00
|
|
|
"/dev/disk/by-path/127.0.0.1:3260-eui.02004567A425678D-lun-0": "/dev/sda",
|
2017-12-13 11:31:38 -05:00
|
|
|
"/dev/disk/by-path/127.0.0.3:3260-eui.03004567A425678D-lun-0": "/dev/sdb",
|
2016-04-19 01:10:00 -04:00
|
|
|
"/dev/dm-2": "/dev/dm-2",
|
|
|
|
|
"/dev/dm-3": "/dev/dm-3",
|
2017-12-13 11:31:38 -05:00
|
|
|
"/dev/sdc": "/dev/sdc",
|
2016-04-19 01:10:00 -04:00
|
|
|
"/dev/sde": "/dev/sde",
|
|
|
|
|
}
|
|
|
|
|
return links[path], nil
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (handler *mockOsIOHandler) WriteFile(filename string, data []byte, perm os.FileMode) error {
|
|
|
|
|
return errors.New("Not Implemented for Mock")
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type fakeFileInfo struct {
|
|
|
|
|
name string
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (fi *fakeFileInfo) Name() string {
|
|
|
|
|
return fi.name
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (fi *fakeFileInfo) Size() int64 {
|
|
|
|
|
return 0
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (fi *fakeFileInfo) Mode() os.FileMode {
|
|
|
|
|
return 777
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (fi *fakeFileInfo) ModTime() time.Time {
|
|
|
|
|
return time.Now()
|
|
|
|
|
}
|
|
|
|
|
func (fi *fakeFileInfo) IsDir() bool {
|
|
|
|
|
return false
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (fi *fakeFileInfo) Sys() interface{} {
|
|
|
|
|
return nil
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func TestFindMultipathDeviceForDevice(t *testing.T) {
|
|
|
|
|
mockDeviceUtil := NewDeviceHandler(&mockOsIOHandler{})
|
|
|
|
|
dev := mockDeviceUtil.FindMultipathDeviceForDevice("/dev/disk/by-path/127.0.0.1:3260-eui.02004567A425678D-lun-0")
|
|
|
|
|
if dev != "/dev/dm-1" {
|
|
|
|
|
t.Fatalf("mpio device not found dm-1 expected got [%s]", dev)
|
|
|
|
|
}
|
|
|
|
|
dev = mockDeviceUtil.FindMultipathDeviceForDevice("/dev/disk/by-path/empty")
|
|
|
|
|
if dev != "" {
|
|
|
|
|
t.Fatalf("mpio device not found '' expected got [%s]", dev)
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2016-07-25 17:34:05 -04:00
|
|
|
func TestFindDeviceForPath(t *testing.T) {
|
2016-04-19 01:10:00 -04:00
|
|
|
io := &mockOsIOHandler{}
|
|
|
|
|
|
|
|
|
|
disk, err := findDeviceForPath("/dev/sde", io)
|
2017-08-25 22:56:47 -04:00
|
|
|
if err != nil {
|
|
|
|
|
t.Fatalf("error finding device for path /dev/sde:%v", err)
|
|
|
|
|
}
|
2017-09-07 15:20:12 -04:00
|
|
|
if disk != "sde" {
|
|
|
|
|
t.Fatalf("disk [%s] didn't match expected sde", disk)
|
|
|
|
|
}
|
2016-04-19 01:10:00 -04:00
|
|
|
disk, err = findDeviceForPath("/returns/a/dev", io)
|
2017-08-25 22:56:47 -04:00
|
|
|
if err != nil {
|
|
|
|
|
t.Fatalf("error finding device for path /returns/a/dev:%v", err)
|
|
|
|
|
}
|
2016-04-19 01:10:00 -04:00
|
|
|
if disk != "sde" {
|
|
|
|
|
t.Fatalf("disk [%s] didn't match expected sde", disk)
|
|
|
|
|
}
|
|
|
|
|
_, err = findDeviceForPath("/returns/non/dev", io)
|
|
|
|
|
if err == nil {
|
|
|
|
|
t.Fatalf("link is to incorrect dev")
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
_, err = findDeviceForPath("/path/doesnt/exist", &osIOHandler{})
|
|
|
|
|
if err == nil {
|
|
|
|
|
t.Fatalf("path shouldn't exist but still doesn't give an error")
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
}
|
2017-12-13 11:31:38 -05:00
|
|
|
|
|
|
|
|
func TestFindSlaveDevicesOnMultipath(t *testing.T) {
|
|
|
|
|
mockDeviceUtil := NewDeviceHandler(&mockOsIOHandler{})
|
|
|
|
|
devices := mockDeviceUtil.FindSlaveDevicesOnMultipath("/dev/dm-1")
|
|
|
|
|
if !reflect.DeepEqual(devices, []string{"/dev/sda", "/dev/sdb"}) {
|
|
|
|
|
t.Fatalf("failed to find devices managed by mpio device. /dev/sda, /dev/sdb expected got [%s]", devices)
|
|
|
|
|
}
|
|
|
|
|
dev := mockDeviceUtil.FindSlaveDevicesOnMultipath("/dev/sdc")
|
|
|
|
|
if len(dev) != 0 {
|
|
|
|
|
t.Fatalf("mpio device not found '' expected got [%s]", dev)
|
|
|
|
|
}
|
|
|
|
|
}
|
Avoid deleted iSCSI LUNs in the kernel
This change ensures that iSCSI block devices are deleted after
unmounting, and implements scanning of individual LUNs rather
than scanning the whole iSCSI bus.
In cases where an iSCSI bus is in use by more than one attachment,
detaching used to leave behind phantom block devices, which could
cause I/O errors, long timeouts, or even corruption in the case
when the underlying LUN number was recycled. This change makes
sure to flush references to the block devices after unmounting.
The original iSCSI code scanned the whole target every time a LUN
was attached. On storage controllers that export multiple LUNs on
the same target IQN, this led to a situation where nodes would
see SCSI disks that they weren't supposed to -- possibly dozens or
hundreds of extra SCSI disks. This caused 3 significant problems:
1) The large number of disks wasted resources on the node and
caused a minor drag on performance.
2) The scanning of all the devices caused a huge number of uevents
from the kernel, causing udev to bog down for multiple minutes in
some cases, triggering timeouts and other transient failures.
3) Because Kubernetes was not tracking all the "extra" LUNs that
got discovered, they would not get cleaned up until the last LUN
on a particular target was detached, causing a logout. This led
to significant complications:
In the time window between when a LUN was unintentially scanned,
and when it was removed due to a logout, if it was deleted on the
backend, a phantom reference remained on the node. In the best
case, the phantom LUN would cause I/O errors and timeouts in the
udev system. In the worst case, the backend could reuse the LUN
number for a new volume, and if that new volume were to be
scheduled to a pod with a phantom reference to the old LUN by the
same number, the initiator could get confused and possibly corrupt
data on that volume.
To avoid these problems, the new implementation only scans for
the specific LUN number it expects to see. It's worth noting that
the default behavior of iscsiadm is to automatically scan the
whole bus on login. That behavior can be disabled by setting
node.session.scan = manual
in iscsid.conf, and for the reasons mentioned above, it is
strongly recommended to set that option. This change still works
regardless of the setting in iscsid.conf, and while automatic
scanning will cause some problems, this change doesn't make the
problems any worse, and can make things better in some cases.
2018-07-24 23:58:19 -04:00
|
|
|
|
|
|
|
|
func TestGetISCSIPortalHostMapForTarget(t *testing.T) {
|
|
|
|
|
mockDeviceUtil := NewDeviceHandler(&mockOsIOHandler{})
|
|
|
|
|
portalHostMap, err := mockDeviceUtil.GetISCSIPortalHostMapForTarget("target1")
|
|
|
|
|
if nil != err {
|
|
|
|
|
t.Fatalf("error getting scsi hosts for target: %v", err)
|
|
|
|
|
}
|
|
|
|
|
if nil == portalHostMap {
|
|
|
|
|
t.Fatal("no portal host map returned")
|
|
|
|
|
}
|
|
|
|
|
if 1 != len(portalHostMap) {
|
|
|
|
|
t.Fatalf("wrong number of map entries in portal host map: %d", len(portalHostMap))
|
|
|
|
|
}
|
|
|
|
|
if 2 != portalHostMap["10.0.0.1:3260"] {
|
|
|
|
|
t.Fatalf("incorrect entry in portal host map: %v", portalHostMap)
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func TestFindDevicesForISCSILun(t *testing.T) {
|
|
|
|
|
mockDeviceUtil := NewDeviceHandler(&mockOsIOHandler{})
|
|
|
|
|
devices, err := mockDeviceUtil.FindDevicesForISCSILun("target1", 1)
|
|
|
|
|
if nil != err {
|
|
|
|
|
t.Fatalf("error getting devices for lun: %v", err)
|
|
|
|
|
}
|
|
|
|
|
if nil == devices {
|
|
|
|
|
t.Fatal("no devices returned")
|
|
|
|
|
}
|
|
|
|
|
if 1 != len(devices) {
|
|
|
|
|
t.Fatalf("wrong number of devices: %d", len(devices))
|
|
|
|
|
}
|
|
|
|
|
if "sda" != devices[0] {
|
|
|
|
|
t.Fatalf("incorrect device %v", devices)
|
|
|
|
|
}
|
|
|
|
|
}
|