Read Write Once Persistent Storage for OpenShift Origin using Gluster

In this blog we shall learn about:

  1. Containers and Persistent Storage
  2. About OpenShift Origin
  3. Terminology and background
  4. Our approach
  5. Setting up
    • Gluster and iSCSI target
    • iSCSI Initiator
    • Origin master and nodes
  6. Conclusion
  7. References

 

Containers and Persistent Storage

As we all know containers are stateless entities which are used to deploy  applications and hence need persistent storage to store  application data for availability across container incarnations.

Persistent storage in containers are of two types, shared and non-shared.
Shared storage:
Consider this as a volume/store where multiple Containers perform both read and write operations on the same data. Useful for applications like web servers that need to serve the same data from multiple container instances.

Non Shared/Read Write Once Storage:
Only a single container can perform write operations to this store at a given time.

This blog will explain about Non Shared Storage for OpenShift Origin using gluster.

 

About OpenShift Origin

OpenShift Origin is a distribution of Kubernetes optimized for continuous application development and multi-tenant deployment.

A few interesting features include Multi-tenancy support, Web console, Centralized administration, capability to automatically deploy applications on a new commit in source repo and etc..

Difference_dock_openshift

Read More @ origin github

 

Terminology and background

Refer to ‘Terminology and background’ section from our previous post

 

Our Approach

With all the background discussed above now I shall jump into actual essence of this blog and explain how we can expose the file in gluster volume as a read write once persistent storage in openshift pods.

The Current version of Kubernetes v1.2.x  which origin uses in my case, does not provide/understand multipathing, this patch got merged in v1.3.alpha3 release

Hence, In this blog I’m going with multipath disabled, once ansible playbook is upgraded to latest origin which use k8s v1.3.0, I shall update the blog to have multipath changes.

In our approach all the OpenShift Origin nodes initiate the iSCSI session, attaches iSCSI target as block device and serve it to pod where the application is running and requires persistent storage.

OpenShiftOrigin

Now without any delay let me walk through the setup details…

 

Setting Up

You need 6 nodes for setting this up, 3 acts as gluster nodes where the iSCSI target is served from and 1 as OpenShift Origin master and other 2 as the iSCSI initiators which also acts as Origin nodes.

  • We create a gluster replica 3 volume using the 3 nodes {Node1, Node2 and Node3}.
  • Define iSCSI target using the same nodes, expose ‘LUN’ from each of them.
  • Use Node 4 and Node 5 as as iSCSI initiators, by logging-in to the iSCSI target session created above (No multipathing)
  • Setup OpenShift Origin cluster by using {Node4, Node5 and Node6}, Node 6 is master and other 2 are slave nodes
  • From Node 6 create the pod and examine the iSCSI target device mount inside it.

Gluster and iSCSI target Setup

Refer to ‘Gluster and iSCSI target Setup’ section from our previous post

iSCSI initiator Setup

Refer to ‘iSCSI initiator Setup’ section from our previous post

OpenShift Origin Master and Nodes Setup

Master -> Node6
Slaves -> Node5 & Node4

Clone the openshift ansible repo
[root@Node6 ~]# git clone https://github.com/openshift/openshift-ansible.git

Install ansible on all the nodes including master
# dnf install -y ansible pyOpenSSL python-cryptography

Configure nodes in inventory file,
all you need to do is replacehost addresses, highlighted in bold
[root@Node6 ~]# cat > /etc/ansible/hosts
# Create an OSEv3 group that contains the masters and nodes groups
[OSEv3:children]
masters
nodes

# Set variables common for all OSEv3 hosts
[OSEv3:vars]
# SSH user, this user should allow ssh based auth without requiring a password
ansible_ssh_user=root

# If ansible_ssh_user is not root, ansible_sudo must be set to true
#ansible_sudo=true

deployment_type=origin

# uncomment the following to enable htpasswd authentication; defaults to DenyAllPasswordIdentityProvider
#openshift_master_identity_providers=[{'name': 'htpasswd_auth', 'login': 'true', 'challenge': 'true',
# 'kind': 'HTPasswdPasswordIdentityProvider', 'filename': '/etc/origin/master/htpasswd'}]

# host group for masters
[masters]
Node6

# host group for nodes, includes region info
[nodes]
Node6 openshift_node_labels="{'region': 'infra', 'zone': 'default'}"
Node5 openshift_node_labels="{'region': 'primary', 'zone': 'east'}"
Node4 openshift_node_labels="{'region': 'primary', 'zone': 'west'}"
^C

Make nodes password less authorise logins, on all machines

Generate ssh key 
# ssh-keygen

Share ssh key with all the nodes, to do so, execute below on master,
$HOSTS being all the addresses/ip including master's, one at a time
[root@Node6 ~]# ssh-copy-id -i ~/.ssh/id_rsa.pub $HOSTS

Just matter of precaution on all the hosts, disable selinux
# setenforce 0

Install some package dependencies, ignored by playbook
[root@Node6 ~]# ansible all -m shell -a "dnf install python2-dnf -y" 
[root@Node6 ~]# ansible all -m shell -a "dnf install python-dbus -y"
[root@Node6 ~]# ansible all -m shell -a "dnf install libsemanage-python -y"

Lets, execute the playbook
[root@Node6 ~]# cd $PATH/openshift-ansible
[root@Node6 openshift-ansible]# ansible-playbook playbooks/byo/config.yml
It takes ~40 minutes to finish this, at least that's what it took me. 

Check all nodes are ready
[root@Node6 ~]# oc get nodes
NAME STATUS AGE
Node4 Ready 1h
Node5 Ready 1h
Node6 Ready,SchedulingDisabled 1h

Check for pods
[root@Node6 ~]# oc get pods

 

login to the origin web console https://Node6:8443
Credentials: user->admin, passwd->admin

Screenshot from 2016-08-16 16-20-01.png

create a New project “say blockstore-gluster”

Attach4.png

 

Switch to 'blockstore-gluster' project
[root@Node6 ~]# oc project blockstore-gluster
Now using project "blockstore-gluster" on server "https://Node6:8443".

Write a manifest/artifact for the pod
[root@Node6 ~]# cat > iscsi-pod.json
{
   "apiVersion": "v1",
   "kind": "Pod",
   "metadata": {
      "name": "glusterpod"
   },
   "spec": {
      "containers": [
         {
            "name": "iscsi-rw",
            "image": "fedora",
            "volumeMounts": [
               {
                  "mountPath": "/mnt/gluster-store",
                  "name": "iscsi-rw"
               }
            ],
            "command": [ "sleep", " 100000" ]
         }
      ],
      "volumes": [
         {
            "name": "iscsi-rw",
            "iscsi": {
               "targetPortal": "Node1:3260",
               "iqn": "iqn.2016-06.org.gluster:Node1",
               "lun": 0,
               "fsType": "xfs",
               "readOnly": false
            }
         }
      ]
   } 
}
^C

Create the pod
[root@Node6 ~]# oc create -f ~/iscsi-pod.json 
pod "glusterpod" created

Get the pod info
[root@Node6 ~]# oc get pods
NAME READY STATUS RESTARTS AGE
glusterpod 0/1 ContainerCreating 0 20s

Check events
[root@Node6 ~]# oc get events -w
FIRSTSEEN LASTSEEN COUNT NAME KIND SUBOBJECT TYPE REASON SOURCE MESSAGE
2016-08-16 16:16:10 +0530 IST 2016-08-16 16:16:10 +0530 IST 1 glusterpod Pod Normal Scheduled {default-scheduler } Successfully assigned glusterpod to dhcp43-73.lab.eng.blr.redhat.com
FIRSTSEEN LASTSEEN COUNT NAME KIND SUBOBJECT TYPE REASON SOURCE MESSAGE
2016-08-16 16:16:14 +0530 IST 2016-08-16 16:16:14 +0530 IST 1 glusterpod Pod spec.containers{iscsi-rw} Normal Pulling {kubelet Node5} pulling image "fedora"
2016-08-16 16:17:17 +0530 IST 2016-08-16 16:17:17 +0530 IST 1 glusterpod Pod spec.containers{iscsi-rw} Normal Pulled {kubelet Node5} Successfully pulled image "fedora"
2016-08-16 16:17:18 +0530 IST 2016-08-16 16:17:18 +0530 IST 1 glusterpod Pod spec.containers{iscsi-rw} Normal Created {kubelet Node5} Created container with docker id 0208911923f1
2016-08-16 16:17:18 +0530 IST 2016-08-16 16:17:18 +0530 IST 1 glusterpod Pod spec.containers{iscsi-rw} Normal Started {kubelet Node5} Started container with docker id 0208911923f1

[root@Node6 ~]# oc get pods
NAME READY STATUS RESTARTS AGE
glusterpod 1/1 Running 0 1m

Get into the pod
[root@Node6 ~]# oc exec -it glusterpod bash

[root@glusterpod /]# df -Th
Filesystem Type Size Used Avail Use% Mounted on
[...]
/dev/sda xfs 8G 33M 8G 1% /mnt/gluster-store
/dev/mapper/fedora_dhcp42--82-root xfs 15G 1.8G 14G 12% /etc/hosts
[...]

[root@glusterpod /]# cd /mnt/gluster-store/
[root@glusterpod gluster-store]# ls
1 10 2 3 4 5 6 7 8 9 


 

Origin Web console with pod running:

Attach1.png

Details of pod:

Attach2.png

That’s cool Isn’t it ?

 

Conclusion

This just showcases how Gluster can be used as a distributed block store with OpenShift Origin cluster. More details about multipathing, integration with Mesos etc. will come by in further posts.

 

References

https://docs.openshift.org/latest/welcome/index.html

https://github.com/openshift/openshift-ansible/

http://kubernetes.io/

http://severalnines.com/blog/installing-kubernetes-cluster-minions-centos7-manage-pods-services

http://rootfs.github.io/iSCSI-Kubernetes/

http://blog.gluster.org/2016/04/using-lio-with-gluster/

https://docs.docker.com/engine/tutorials/dockervolumes/http://scst.sourceforge.net/scstvslio.html

http://events.linuxfoundation.org/sites/events/files/slides/tcmu-bobw_0.pdf

https://www.kernel.org/doc/Documentation/target/tcmu-design.txt

https://lwn.net/Articles/424004/

http://www.gluster.org/community/documentation/index.php/GlusterFS_Documentation

 

Advertisements

Gluster Solution for Non Shared Persistent Storage in Docker Container

In this blog we shall see

  1. Containers and Persistent Storage
  2. Terminology and background
  3. Our approach
  4. Setting up
    • Gluster and iSCSI target
    • iSCSI Initiator
    • Docker host and Container
  5. Conclusion
  6. References

Containers and Persistent Storage

As we all know containers are stateless entities which are used to deploy  applications and hence need persistent storage to store  application data for availability across container incarnations.

Persistent storage in containers are of two types, shared and non-shared.
Shared storage:
Consider this as a volume/store where multiple Containers perform both read and write operations on the same data. Useful for applications like web servers that need to serve the same data from multiple container instances.

Non Shared Storage:
Only a single container can perform write operations to this store at a given time.

This blog will explain about Non Shared Storage for containers using gluster.

Terminology and background

Gluster is a well known scale-out distributed storage system, flexible in its design and to use. One of its key goals is to provide high availability of data.  Despite its distributed nature, Gluster is very easy to set up and use. Addition and removal of storage servers from a Gluster cluster is very easy. These capabilities along with other data services that Gluster provides makes it a very nice software defined storage platform.

We can access glusterfs via FUSE module. However to perform a single filesystem operation various context switches are required which leads to performance issues. Libgfapi is a userspace library for accessing data in Glusterfs. It can perform IO on gluster volumes without the FUSE module, kernel VFS layer and hence requires no context switches. It exposes a filesystem like API for accessing gluster volumes. Samba, NFS-Ganesha, QEMU and now the tcmu-runner all use libgfapi to integrate with Glusterfs.

The SCSI subsystem uses a sort of client-server model.  The Client/Initiator request IO happen through target which is a storage device. The SCSI target subsystem enables a computer node to behave as a SCSI storage device, responding to storage requests by other SCSI initiator nodes.

In simple terms SCSI is a set of standards for physically connecting and transferring data between computers and peripheral devices.

The most common implementation of the SCSI target subsystem is an iSCSI server, iSCSI transports block level data between the iSCSI initiator and the target which resides on the actual storage device. iSCSi protocol wraps up the SCSI commands and sends it over TCP/IP layer. Up on receiving the packets at the other end it disassembles them to form the same SCSI commands, hence on the OS’es it seen as local SCSI device.

In other words iSCSI is SCSI over TCP/IP.

The LIO project began with the iSCSI design as its core objective, and created a generic SCSI target subsystem to support iSCSI. LIO is the SCSI target in the Linux kernel. It is entirely kernel code, and allows exported SCSI logical units (LUNs) to be backed by regular files or block devices.

TCM is another name for LIO, an in-kernel iSCSI target (server). As we know existing TCM targets run in the kernel. TCMU (TCM in Userspace) allows userspace programs to be written which act as iSCSI targets. These enables wider variety of backstores without kernel code. Hence the TCMU userspace-passthrough backstore allows a userspace process to handle requests to a LUN.

One such backstore with best clustered network storage capabilities is GlusterFS

Tcmu-runner utilizes the TCMU framework handling the messy details of the TCMU interface, allows target in the glusterfs volume to be used over gluster’s libgfapi interface.

TargetCli is the general management platform for the LIO/TCM/TCMU. TargetCli with its shell interface is used to configure LIO.

Our Approach

With all the background discussed above now I shall jump into actual essence of this blog and explain how we can expose the file in gluster volume as a nonshared persistant storage in Docker

We use iSCSI which are highly sensitive to network performance, jitters in the connections will cause iSCSI connection to perform poorly.

Also we all know that container suffer from suboptimal network performance, because of the fact that Docker NAT doesn’t deliver good performance.

Hence in our approach Docker initiates the iSCSI session, attaches iSCSI target as block device with multipath enabled, mounts it in local directory, and share that via bind mount to the container. This approach doesn’t need Docker NAT (hence we don’t lose performance).

imageedit_1_7869107488

Now without any delay let me walk through the setup details…

Setting Up

You need 4 nodes for setting this up, 3 acts as gluster nodes where the iSCSI target is served from and 1 machine as a iSCSI Initiator/Docker Host where container deployment happens.

Gluster and iSCSI target Setup

On all the nodes I have Installed freshly baked fedora 24 Beta server.

Perform below on all the 3 gluster nodes:

# dnf upgrade
# dnf install glusterfs-server
got 3.8.0-0.2.rc2.fc24.x86_64
# iptables -F
# vi /etc/glusterfs/glusterd.vol
add 'option rpc-auth-allow-insecure on' this is needed by libgfapi
# systemctl start glusterd

On Node1 perform below:

# form a gluster trusted pool of nodes
[root@Node1 ~]# gluster peer probe Node2
[root@Node1 ~]# gluster peer probe Node3
[root@Node1 ~]# gluster pool list
UUID                                    Hostname        State
51023fac-edc7-4149-9e3c-6b8207a02f7e    Node1   *Connected 
ed9af9d6-21f0-4a37-be2c-5c23eff4497e    Node2    Connected 
052e681e-cdc8-4d3b-87a4-39de26371b0f    Node3    Connected
 `
# create volume
[root@Node1 ~]# gluster vol create nonshared-store replica 3 Node1:/subvol1 Node2:/subvol2 Node3:/subvol3 force

# volume set to allow insecure port ranges
[root@Node1 ~]# gluster vol set nonshared-store server.allow-insecure on

# start the volume 
[root@Node1 ~]# gluster vol start nonshared-store
[root@Node1 ~]# gluster vol status
check the status

# mount volume and create a required target file of 8G in volume
[root@Node1 ~]# mount -t glusterfs Node1:/nonshared-store /mnt
[root@Node1 ~]# cd /mnt
[root@Node1 ~]# fallocate -l 8G app-store.img

# finally unmount the volume
[root@Node1 ~]# umount /mnt

Now we have done with the gluster  setup, lets expose file ‘app-store.img’ as an iSCSI target by creating a LUN

Again On Node1:

[root@Node1 ~]# dnf install tcmu-runner targetcli

# enter the admin console
[root@Node1 ~]# targetcli
targetcli shell version 2.1.fb42
Copyright 2011-2013 by Datera, Inc and others.
For help on commands, type 'help'.
/> ls
o- / ...................................................... [...]
  o- backstores ........................................... [...]
  | o- block ............................... [Storage Objects: 0]
  | o- fileio .............................. [Storage Objects: 0]
  | o- pscsi ............................... [Storage Objects: 0]
  | o- ramdisk ............................. [Storage Objects: 0]
  | o- user:glfs ........................... [Storage Objects: 0]
  | o- user:qcow ........................... [Storage Objects: 0]
  o- iscsi ......................................... [Targets: 0]
  o- loopback ...................................... [Targets: 0]
  o- vhost ......................................... [Targets: 0]

# Create a Image file with the name "glfsLUN" on gluster volume with 8G
/> cd /backstores/user:glfs
/backstores/user:glfs> create glfsLUN 8G nonshared-store@Node1/app-store.img
Created user-backed storage object glfsLUN size 8589934592.

# create a target
/backstores/user:glfs> cd /iscsi/
/iscsi> create iqn.2016-06.org.gluster:Node1            
Created target iqn.2016-06.org.gluster:Node1.
Created TPG 1.
Global pref auto_add_default_portal=true
Created default portal listening on all IPs (0.0.0.0), port 3260.

# set LUN
/iscsi> cd /iscsi/iqn.2016-06.org.gluster:Node1/tpg1/luns
iscsi/iqn.20...de1/tpg1/luns> create /backstores/user:glfs/glfsLUN 0
Created LUN 0.
iscsi/iqn.20...de1/tpg1/luns> cd /

# set ACL (it's the IQN of an initiator you permit to connect)
# Copy InitiatorName from Docker host machine
# which will be in ‘/etc/iscsi/initiatorname.iscsi’
# In my case it is the one put in bold at next commands
/> cd /iscsi/iqn.2016-06.org.gluster:Node1/tpg1/acls 
/iscsi/iqn.20...de1/tpg1/acls> create iqn.1994-05.com.redhat:8277148d16b2
Created Node ACL for iqn.1994-05.com.redhat:8277148d16b2
Created mapped LUN 0.

# set UserID and password for authentication
/iscsi/iqn.20...de1/tpg1/acls> cd iqn.1994-05.com.redhat:8277148d16b2
/iscsi/iqn.20...:8277148d16b2> set auth userid=username 
Parameter userid is now 'username'.
/iscsi/iqn.20...:8277148d16b2> set auth password=password 
Parameter password is now 'password'.
/iscsi/iqn.20...:8277148d16b2> cd /

/> saveconfig
Last 10 configs saved in /etc/target/backup.
Configuration saved to /etc/target/saveconfig.json

/> exit
Global pref auto_save_on_exit=true
Last 10 configs saved in /etc/target/backup.
Configuration saved to /etc/target/saveconfig.json

Multipathing will help in achieving high availability of the LUN at client side, so even when a node is down for maintenance your LUN remains accessible.

Using the same wwn across the nodes will help enable multipathing, we need to ensure the LUN exported by all the three gateways/nodes to share the same wwn – if they don’t match, the client will see three devices, not three paths to the same device.

Copy the wwn from the Node1, the one shown in bold below:

[root@Node1 ~]# cat /etc/target/saveconfig.json
{
 "fabric_modules": [],
 "storage_objects": [
 {
 "config": "glfs/nonshared-store@Node1/app-store.img",
 "name": "glfsLUN",
 "plugin": "user",
 "size": 8589934592,
 "wwn": "cdc1e292-c21a-41ce-aa3f-d49658633bdf"
 } 
],
"targets": [
[...]

Now On Node2:

[root@Node2 ~]# dnf install tcmu-runner targetcli

# enter the admin console
[root@Node2 ~]# targetcli
targetcli shell version 2.1.fb42
Copyright 2011-2013 by Datera, Inc and others.
For help on commands, type 'help'.
/> ls
o- / ...................................................... [...]
  o- backstores ........................................... [...]
  | o- block ............................... [Storage Objects: 0]
  | o- fileio .............................. [Storage Objects: 0]
  | o- pscsi ............................... [Storage Objects: 0]
  | o- ramdisk ............................. [Storage Objects: 0]
  | o- user:glfs ........................... [Storage Objects: 0]
  | o- user:qcow ........................... [Storage Objects: 0]
  o- iscsi ......................................... [Targets: 0]
  o- loopback ...................................... [Targets: 0]
  o- vhost ......................................... [Targets: 0]

# create a Image file with the name "glfsLUN" on gluster volume with 8G
/> /backstores/user:glfs create glfsLUN 8G nonshared-store@Node2/app-store.img
Created user-backed storage object glfsLUN size 8589934592.

/> saveconfig
Last 10 configs saved in /etc/target/backup.
Configuration saved to /etc/target/saveconfig.json

/> exit
Global pref auto_save_on_exit=true
Last 10 configs saved in /etc/target/backup.
Configuration saved to /etc/target/saveconfig.json

[root@Node2 ~]# vi /etc/target/saveconfig.json
edit wwn to point to one that is copied from Node1 and save

[root@Node2 ~]# systemctl restart target
[root@Node2 ~]# targetcli
targetcli shell version 2.1.fb42
Copyright 2011-2013 by Datera, Inc and others.
For help on commands, type 'help'.
# create a target
/> /iscsi/ create iqn.2016-06.org.gluster:Node2            
Created target iqn.2016-06.org.gluster:Node2.
Created TPG 1.
Global pref auto_add_default_portal=true
Created default portal listening on all IPs (0.0.0.0), port 3260.

# set LUN
/> cd /iscsi/iqn.2016-06.org.gluster:Node2/tpg1/luns
iscsi/iqn.20...de2/tpg1/luns> create /backstores/user:glfs/glfsLUN 0
Created LUN 0.
iscsi/iqn.20...de2/tpg1/luns> cd /

# set ACL (it's the IQN of an initiator you permit to connect)
# Copy InitiatorName from Docker host machine
# which will be in ‘/etc/iscsi/initiatorname.iscsi’
# In my case it is the one put in bold at next commands
/> cd /iscsi/iqn.2016-06.org.gluster:Node2/tpg1/acls 
/iscsi/iqn.20...de2/tpg1/acls> create iqn.1994-05.com.redhat:8277148d16b2
Created Node ACL for iqn.1994-05.com.redhat:8277148d16b2
Created mapped LUN 0.

# set UserID and password for authentication
/iscsi/iqn.20...de2/tpg1/acls> cd iqn.1994-05.com.redhat:8277148d16b2
/iscsi/iqn.20...:8277148d16b2> set auth userid=username 
Parameter userid is now 'username'.
/iscsi/iqn.20...:8277148d16b2> set auth password=password 
Parameter password is now 'password'.
/iscsi/iqn.20...:8277148d16b2> cd /

/> saveconfig
Last 10 configs saved in /etc/target/backup.
Configuration saved to /etc/target/saveconfig.json

/> exit
Global pref auto_save_on_exit=true
Last 10 configs saved in /etc/target/backup.
Configuration saved to /etc/target/saveconfig.json

Note: Replicate Node2 setup on Node3 as well, avoiding the duplication

Setting up iSCSI initiator

On the fourth Machine:

[root@DkNode ~]# dnf install iscsi-initiator-utils sg3_utils 

# Multipathing to achieve high availability
[root@DkNode ~]# mpathconf 
multipath is enabled
find_multipaths is enabled
user_friendly_names is enabled
dm_multipath module is not loaded
multipathd is not running

[root@DkNode ~]# modprobe dm_multipath
[root@DkNode ~]# lsmod | grep dm_multipath
dm_multipath           24576  0

[root@DkNode ~]# cat /etc/multipath.conf
cat: /etc/multipath.conf: No such file or directory

[root@DkNode ~]# mpathconf --enable
[root@DkNode ~]# cat >> /etc/multipath.conf                                                                                     

# LIO iSCSI
devices { 
        device { 
                vendor "LIO-ORG"
                user_friendly_names "yes" # names like mpatha
                path_grouping_policy "failover" # one path per group
                path_selector "round-robin 0"
                path_checker "tur"
                prio "const"
                rr_weight "uniform"
        } 
}
^C

[root@DkNode ~]# systemctl start multipathd
[root@DkNode ~]# mpathconf 
multipath is enabled
find_multipaths is enabled
user_friendly_names is enabled
dm_multipath module is loaded
multipathd is running

[root@DkNode ~]# vi /etc/iscsi/iscsid.conf
# uncomment below line
node.session.auth.authmethod = CHAP
[...]
# uncomment and edit username and password if required
node.session.auth.username = username
node.session.auth.password = password

[root@DkNode ~]# systemctl restart iscsid 
[root@DkNode ~]# systemctl enable iscsid
# One way of Login to the iSCSI target
[root@DkNode ~]#  iscsiadm -m discovery -t st -p Node1 -l 
Node1:3260,1 iqn.2016-06.org.gluster:Node1
Logging in to [iface: default, target: iqn.2016-06.org.gluster:Node1, portal: Node1,3260] (multiple)
Login to [iface: default, target: iqn.2016-06.org.gluster:Node1, portal: Node1,3260] successful.

[root@DkNode ~]#  iscsiadm -m discovery -t st -p Node2 -l 
Node2:3260,1 iqn.2016-06.org.gluster:Node2
Logging in to [iface: default, target: iqn.2016-06.org.gluster:Node2, portal: Node2,3260] (multiple)
Login to [iface: default, target: iqn.2016-06.org.gluster:Node2, portal: Node2,3260] successful.

[root@DkNode ~]#  iscsiadm -m discovery -t st -p Node3 -l 
Node3:3260,1 iqn.2016-06.org.gluster:Node3
Logging in to [iface: default, target: iqn.2016-06.org.gluster:Node3, portal: Node3,3260] (multiple)
Login to [iface: default, target: iqn.2016-06.org.gluster:Node3, portal: Node3,3260] successful.

# Here you can see three paths to the same device
[root@DkNode ~]# multipath -ll
mpatha (36001405cdc1e292c21a41ceaa3fd4965) dm-2 LIO-ORG ,TCMU device     
size=8.0G features='0' hwhandler='0' wp=rw
`-+- policy='queue-length 0' prio=1 status=active
  |- 2:0:0:0 sda 8:0  active ready running
  |- 3:0:0:0 sdb 8:16 active ready running
  `- 4:0:0:0 sdc 8:32 active ready running

[root@DkNode ~]# lsblk
NAME                       MAJ:MIN RM  SIZE RO TYPE  MOUNTPOINT
sda                          8:0    0    8G  0 disk  
└─mpatha                   253:2    0    8G  0 mpath 
sdb                          8:16   0    8G  0 disk  
└─mpatha                   253:2    0    8G  0 mpath 
sdc                          8:32   0    8G  0 disk  
└─mpatha                   253:2    0    8G  0 mpath 
[...]

Some ways to check multipathing is working as expected

# From sda, sdb, sdc check wwn's are same and Nodes are as expected

[root@DkNode ~]# sg_inq -i /dev/sda 
VPD INQUIRY: Device Identification page
  Designation descriptor number 1, descriptor length: 49
    designator_type: T10 vendor identification,  code_set: ASCII
    associated with the addressed logical unit
      vendor id: LIO-ORG 
      vendor specific: cdc1e292-c21a-41ce-aa3f-d49658633bdf
  Designation descriptor number 2, descriptor length: 20
    designator_type: NAA,  code_set: Binary
    associated with the addressed logical unit
      NAA 6, IEEE Company_id: 0x1405
      Vendor Specific Identifier: 0xcdc1e292c
      Vendor Specific Identifier Extension: 0x21a41ceaa3fd4965
      [0x6001405cdc1e292c21a41ceaa3fd4965]
  Designation descriptor number 3, descriptor length: 51
    designator_type: vendor specific [0x0],  code_set: ASCII
    associated with the addressed logical unit
      vendor specific: glfs/nonshared-store@Node1/app-store.img

[root@DkNode ~]# sg_inq -i /dev/sdb
VPD INQUIRY: Device Identification page
  Designation descriptor number 1, descriptor length: 49
    designator_type: T10 vendor identification,  code_set: ASCII
    associated with the addressed logical unit
      vendor id: LIO-ORG 
      vendor specific: cdc1e292-c21a-41ce-aa3f-d49658633bdf
  Designation descriptor number 2, descriptor length: 20
    designator_type: NAA,  code_set: Binary
    associated with the addressed logical unit
      NAA 6, IEEE Company_id: 0x1405
      Vendor Specific Identifier: 0xcdc1e292c
      Vendor Specific Identifier Extension: 0x21a41ceaa3fd4965
      [0x6001405cdc1e292c21a41ceaa3fd4965]
  Designation descriptor number 3, descriptor length: 52
    designator_type: vendor specific [0x0],  code_set: ASCII
    associated with the addressed logical unit
      vendor specific: glfs/nonshared-store@Node2/app-store.img

[root@DkNode ~]# sg_inq -i /dev/sdc
VPD INQUIRY: Device Identification page
  Designation descriptor number 1, descriptor length: 49
    designator_type: T10 vendor identification,  code_set: ASCII
    associated with the addressed logical unit
      vendor id: LIO-ORG 
      vendor specific: cdc1e292-c21a-41ce-aa3f-d49658633bdf
  Designation descriptor number 2, descriptor length: 20
    designator_type: NAA,  code_set: Binary
    associated with the addressed logical unit
      NAA 6, IEEE Company_id: 0x1405
      Vendor Specific Identifier: 0xcdc1e292c
      Vendor Specific Identifier Extension: 0x21a41ceaa3fd4965
      [0x6001405cdc1e292c21a41ceaa3fd4965]
  Designation descriptor number 3, descriptor length: 51
    designator_type: vendor specific [0x0],  code_set: ASCII
    associated with the addressed logical unit
      vendor specific: glfs/nonshared-store@Node3/app-store.img

[root@DkNode ~]# ls -l /dev/mapper/
total 0
[...]
lrwxrwxrwx. 1 root root       7 Jun 21 12:13 mpatha -> ../dm-2

[root@DkNode ~]# sg_inq -i /dev/mapper/mpatha 
VPD INQUIRY: Device Identification page
  Designation descriptor number 1, descriptor length: 49
    designator_type: T10 vendor identification,  code_set: ASCII
    associated with the addressed logical unit
      vendor id: LIO-ORG 
      vendor specific: cdc1e292-c21a-41ce-aa3f-d49658633bdf
  Designation descriptor number 2, descriptor length: 20
    designator_type: NAA,  code_set: Binary
    associated with the addressed logical unit
      NAA 6, IEEE Company_id: 0x1405
      Vendor Specific Identifier: 0xcdc1e292c
      Vendor Specific Identifier Extension: 0x21a41ceaa3fd4965
      [0x6001405cdc1e292c21a41ceaa3fd4965]
  Designation descriptor number 3, descriptor length: 51
    designator_type: vendor specific [0x0],  code_set: ASCII
    associated with the addressed logical unit
      vendor specific: glfs/nonshared-store@Node3/app-store.img

Here we partition the target device and mount it local to iSCSI initiator/Docker Host

# partition the disk file
[root@DkNode ~]# sgdisk -n 1:2048 /dev/mapper/mpatha 
Creating new GPT entries.
Warning: The kernel is still using the old partition table.
The new table will be used at the next reboot or after you
run partprobe(8) or kpartx(8)
The operation has completed successfully.

# now commit the changes to disk
[root@DkNode ~]# partprobe /dev/mapper/mpatha

# check if that appeared as expected
[root@DkNode ~]# lsblk
NAME                       MAJ:MIN RM  SIZE RO TYPE  MOUNTPOINT
sda                          8:0    0    8G  0 disk  
└─mpatha                   253:2    0    8G  0 mpath 
  └─mpatha1                253:3    0    8G  0 part  
sdb                          8:16   0    8G  0 disk  
└─mpatha                   253:2    0    8G  0 mpath 
  └─mpatha1                253:3    0    8G  0 part  
sdc                          8:32   0    8G  0 disk  
└─mpatha                   253:2    0    8G  0 mpath 
  └─mpatha1                253:3    0    8G  0 part  
[...]

[root@DkNode ~]# sgdisk -p /dev/mapper/mpatha
Disk /dev/mapper/mpatha: 16777216 sectors, 8.0 GiB
Logical sector size: 512 bytes
Disk identifier (GUID): 71183D83-4290-41F4-8EF9-69B3D14495F8
Partition table holds up to 128 entries
First usable sector is 34, last usable sector is 16777182
Partitions will be aligned on 2048-sector boundaries
Total free space is 2014 sectors (1007.0 KiB)
Number  Start (sector)    End (sector)  Size       Code  Name
   1            2048        16777182   8.0 GiB     8300  

[root@DkNode ~]# ls -l /dev/mapper/
total 0
[...]
lrwxrwxrwx. 1 root root       7 Jun 21 12:23 mpatha -> ../dm-2
lrwxrwxrwx. 1 root root       7 Jun 21 12:23 mpatha1 -> ../dm-3

# format the target device
[root@DkNode ~]#  mkfs.xfs /dev/mapper/mpatha1 
meta-data=/dev/mapper/mpatha1    isize=512    agcount=4, agsize=524223 blks
         =                       sectsz=512   attr=2, projid32bit=1
         =                       crc=1        finobt=1, sparse=0
data     =                       bsize=4096   blocks=2096891, imaxpct=25
         =                       sunit=0      swidth=0 blks
naming   =version 2              bsize=4096   ascii-ci=0 ftype=1
log      =internal log           bsize=4096   blocks=2560, version=2
         =                       sectsz=512   sunit=0 blks, lazy-count=1
realtime =none                   extsz=4096   blocks=0, rtextents=0

# mount the iSCSI target

[root@DkNode ~]# mkdir /root/nonshared-store/

[root@DkNode ~]# mount /dev/mapper/mpatha1 /root/nonshared-store/

[root@DkNode ~]# df -Th
Filesystem                         Type      Size  Used Avail Use% Mounted on
[...]
/dev/mapper/mpatha1                xfs       8.0G   33M  8.0G   1% /root/nonshared-store/

Let me point you, the iSCSI initiator for us will be the Docker host and viz.

Setting up Docker host and container

[root@DkNode ~]# touch /root/nonshared-store/{1..10}

[root@DkNode ~]# ls /root/nonshared-store/
1 10 2 3 4 5 6 7 8 9

[root@DkNode ~]# cat > /etc/yum.repos.d/docker.repo
[dockerrepo]
name=Docker Repository
baseurl=https://yum.dockerproject.org/repo/main/fedora/$releasever/
enabled=1
gpgcheck=1
gpgkey=https://yum.dockerproject.org/gpg
^c

[root@DkNode ~]# dnf install docker-engine

[root@DkNode ~]# systemctl start docker

# create a container
[root@DkNode ~]# docker run --name bindmount -v /root/nonshared-store/:/mnt:z -t -i fedora /bin/bash

--name         Assign a name to the container
-v             Create a bind mount. If you specify [[HOST-DIR:]CONTAINER-DIR[:OPTIONS]]
               Option: z is the selinux label (instead of ':z' you are free to do # setenforce 0) 
-t             Allocate a pseudo-TTY
-i             Keep STDIN open (Interactive)

# docker interactive tty is here for us
[root@5bbb1e4cb8f8 /]# ls /mnt/
1 10 2 3 4 5 6 7 8 9

[root@5bbb1e4cb8f8 /]# df -Th
Filesystem                               Type   Size  Used Avail Use% Mounted on
[...]
/dev/mapper/mpatha1                      xfs    8.0G   33M  8.0G   1% /mnt
/dev/mapper/fedora_dhcp42--88-root       xfs     15G  1.2G   14G   8% /etc/hosts
shm                                      tmpfs   64M     0   64M   0% /dev/sh

Conclusion

This just showcases how Gluster can be used as a distributed block store for containers. More details about high availability, integration with Kubernetes etc. will come by in further posts.

References

http://domino.research.ibm.com/library/cyberdig.nsf/papers/0929052195DD819C85257D2300681E7B/$File/rc25482.pdf

http://rootfs.github.io/iSCSI-Kubernetes/

http://blog.gluster.org/2016/04/using-lio-with-gluster/

https://docs.docker.com/engine/tutorials/dockervolumes/http://scst.sourceforge.net/scstvslio.html

http://events.linuxfoundation.org/sites/events/files/slides/tcmu-bobw_0.pdf

https://www.kernel.org/doc/Documentation/target/tcmu-design.txt

https://lwn.net/Articles/424004/

http://www.gluster.org/community/documentation/index.php/GlusterFS_Documentation