GlusterFS for SysAdmins
Justin Clift, GlusterFS Integration
Open Source and Standards team @ Red Hat
2013-04
#whoami
● Experienced SysAdmin (Solaris and
Linux) for many, many years
● Mostly worked on Mission Critical
systems in corporate/enterprise
environments (eg. Telco, banking,
Insurance)
● Has been helping build Open Source
years (PostgreSQL, OpenOffice)
● Dislikes networks being a bottleneck
(likes Infiniband. A lot :>)
● Joined Red Hat mid 2010
Agenda
● Technology Overview
● Scaling Up and Out
● A Peek at GlusterFS Logic
● Redundancy and Fault Tolerance
● Data Access
● General Administration
● Use Cases
● Common Pitfalls
Technology
Overview
GlusterFS for SysAdmins
What is GlusterFS?
● POSIX-Like Distributed File System
● No Metadata Server
● Network Attached Storage (NAS)
● Heterogeneous Commodity Hardware
● Aggregated Storage and Memory
● Standards-Based – Clients, Applications, Networks
● Flexible and Agile Scaling
● Capacity – Petabytes and beyond
● Performance – Thousands of Clients
● Single Global Namespace
GlusterFS vs. Traditional Solutions
● A basic NAS has limited scalability and redundancy
● Other distributed filesystems limited by metadata
● SAN is costly & complicated but high performance &
scalable
● GlusterFS
● Linear Scaling
● Minimal Overhead
● High Redundancy
● Simple and Inexpensive Deployment
Technology
Stack
GlusterFS for SysAdmins
Terminology
● Brick
● A filesystem mountpoint
● A unit of storage used as a GlusterFS building block
● Translator
● Logic between the bits and the Global Namespace
● Layered to provide GlusterFS functionality
● Volume
● Bricks combined and passed through translators
● Node / Peer
● Server running the gluster daemon and sharing
volumes
Foundation Components
● Private Cloud (Datacenter)
● Common Commodity x86_64 Servers
● Public Cloud
● Amazon Web Services (AWS)
● EC2 + EBS
Disk, LVM, and Filesystems
● Direct-Attached Storage (DAS)
-or-
● Just a Bunch Of Disks (JBOD)
● Hardware RAID
● Logical Volume Management (LVM)
● XFS, EXT3/4, BTRFS
● Extended attributes (xattr's) support required
● XFS is strongly recommended for GlusterFS
Gluster Components
● glusterd
● Elastic volume management daemon
● Runs on all export servers
● Interfaced through gluster CLI
● glusterfsd
● GlusterFS brick daemon
● One process for each brick
● Managed by glusterd
Gluster Components
● glusterfs
● NFS server daemon
● FUSE client daemon
● mount.glusterfs
● FUSE native mount tool
● gluster
● Gluster Console Manager (CLI)
Data Access Overview
● GlusterFS Native Client
● Filesystem in Userspace (FUSE)
● NFS
● Built-in Service
● SMB/CIFS
● Samba server required
● Unified File and Object (UFO)
● Simultaneous object-based access
Putting it All Together
Scaling
GlusterFS for SysAdmins
Scaling Up
● Add disks and filesystems to a node
● Expand a GlusterFS volume by adding bricks
XFS
Scaling Out
● Add GlusterFS nodes to trusted pool
● Add filesystems as new bricks
Under
the Hood
GlusterFS for SysAdmins
Elastic Hash Algorithm
● No central metadata
● No Performance Bottleneck
● Eliminates risk scenarios
● Location hashed intelligently on path and filename
● Unique identifiers, similar to md5sum
● The “Elastic” Part
● Files assigned to virtual volumes
● Virtual volumes assigned to multiple bricks
● Volumes easily reassigned on the fly
Translators
Distribution
and Replication
GlusterFS for SysAdmins
Distributed Volume
● Files “evenly” spread across bricks
● File-level RAID 0
● Server/Disk failure could be catastrophic
Replicated Volume
● Copies files to multiple bricks
● File-level RAID 1
Distributed Replicated Volume
● Distributes files across replicated bricks
● RAID 1 plus improved read performance
Geo Replication
● Asynchronous across LAN, WAN, or Internet
● Master-Slave model -- Cascading possible
● Continuous and incremental
● Data is passed between defined master and slave only
Replicated Volumes vs Geo-replication
Replicated Volumes Geo-replication
Mirrors data across clusters Mirrors data across geographically
distributed clusters
Provides high-availability Ensures backing up of data for disaster
recovery
Synchronous replication (each and every Asynchronous replication (checks for the
file operation is sent across all the bricks) changes in files periodically and syncs
them on detecting differences)
Layered
Functionality
GlusterFS for SysAdmins
Striped Volumes
● Individual files split among bricks
● Similar to RAID 0
● Limited Use Cases – HPC Pre/Post Processing
Distributed Striped Volume
● Files striped across two or more nodes
● Striping plus scalability
Striped Replicated Volume
● GlusterFS 3.3+
● Similar to RAID 10 (1+0)
Distributed Striped Replicated Volume
● GlusterFS 3.3+
● Limited Use Cases – Map Reduce
Data Access
GlusterFS for SysAdmins
GlusterFS Native Client (FUSE)
● FUSE kernel module allows the filesystem to be built
and operated entirely in userspace
● Specify mount to any GlusterFS node
● Native Client fetches volfile from mount server, then
communicates directly with all nodes to access data
● Recommended for high concurrency and high write
performance
● Load is inherently balanced across distributed volumes
NFS
● Standard NFS v3 clients
● Note: Mount with vers=3 option
● Standard automounter is supported
● Mount to any node, or use a load balancer
● GlusterFS NFS server includes Network Lock Manager
(NLM) to synchronize locks across clients
● Better performance for reading many small files from a
single client
● Load balancing must be managed externally
SMB/CIFS
● GlusterFS volume is first mounted with the Native
Client
● Redundantly on the GlusterFS peer
-or-
● On an external server
● Native mount point is then shared via Samba
● Must be setup on each node you wish to connect to via
CIFS
● Load balancing must be managed externally
General
Administration
GlusterFS for SysAdmins
Preparing a Brick
# lvcreate -L 100G -n lv_brick1 vg_server1
# mkfs -t xfs -i size=512 /dev/vg_server1/lv_brick1
# mkdir /brick1
# mount /dev/vg_server1/lv_brick1 /brick1
# echo '/dev/vg_server1/lv_brick1 /brick1 xfs defaults 1 2' >> /etc/fstab
Adding Nodes (peers) and Volumes
Peer Probe
gluster> peer probe server3
gluster> peer status
Number of Peers: 2
Hostname: server2
Uuid: 5e987bda-16dd-43c2-835b-08b7d55e94e5
State: Peer in Cluster (Connected)
Hostname: server3
Uuid: 1e0ca3aa-9ef7-4f66-8f15-cbc348f29ff7
State: Peer in Cluster (Connected)
Distributed Volume
gluster> volume create my-dist-vol server2:/brick2 server3:/brick3
gluster> volume info my-dist-vol
Volume Name: my-dist-vol
Type: Distribute
Status: Created
Number of Bricks: 2
Transport-type: tcp
Bricks:
Brick1: server2:/brick2
Brick2: server3:/brick3
gluster> volume start my-dist-vol
Distributed Striped Replicated Volume
gluster> volume create test-volume replica 2 stripe 2 transport tcp \
server1:/exp1 server1:/exp2 server2:/exp3 server2:/exp4 \
server3:/exp5 server3:/exp6 server4:/exp7 server4:/exp8
Multiple bricks of a replicate volume are present on the same server. This setup is not
optimal.
Do you still want to continue creating the volume? (y/n) y
Creation of volume test-volume has been successful. Please start the volume to access
data.
<-
<- test-volume
test-volume
replica
replica 22
stripe
stripe 22
<-
<- distributed
distributed files
files ->
->
Distributed Striped Replicated Volume
gluster> volume create test-volume stripe 2 replica 2 transport tcp \
server1:/exp1 server2:/exp3 server1:/exp2 server2:/exp4 \
server3:/exp5 server4:/exp7 server3:/exp6 server4:/exp8
Creation of volume test-volume has been successful. Please start the volume to access
data.
gluster> volume info test-volume
Volume Name: test-volume
Type: Distributed-Striped-Replicate
Volume ID: 8f8b8b59-d1a1-42fe-ae05-abe2537d0e2d
Status: Created
Number of Bricks: 2 x 2 x 2 = 8
Transport-type: tcp
Bricks:
Brick1: server1:/exp1
Brick2: server2:/exp3
Brick3: server1:/exp2
Brick4: server2:/exp4
Brick5: server3:/exp5
Brick6: server4:/exp7
Brick7: server3:/exp6
Brick8: server4:/exp8
Manipulating Bricks in a Volume
gluster> volume add-brick my-dist-vol server4:/brick4
gluster> volume rebalance my-dist-vol fix-layout start
gluster> volume rebalance my-dist-vol start
gluster> volume rebalance my-dist-vol status
Node Rebalanced-files size scanned failures status
--------- ----------- ----------- ----------- ----------- ------------
localhost 112 15674 170 0 completed
10.16.156.72 140 3423 321 2 completed
gluster> volume remove-brick my-dist-vol server2:/brick2 start
gluster> volume remove-brick my-dist-vol server2:/brick2 status
Node Rebalanced-files size scanned failures status
--------- ----------- ----------- ----------- ----------- ------------
localhost 16 16777216 52 0 in progress
192.168.1.1 13 16723211 47 0 in progress
gluster> volume remove-brick my-dist-vol server2:/brick2 commit
Migrating Data / Replacing Bricks
gluster> volume replace-brick my-dist-vol server3:/brick3 server5:/brick5 start
gluster> volume replace-brick my-dist-vol server3:/brick3 server5:/brick5 status
Current File = /usr/src/linux-headers-2.6.31-14/block/Makefile
Number of files migrated = 10567
Migration complete
gluster> volume replace-brick my-dist-vol server3:/brick3 server5:/brick5 commit
Volume Options
Auth
gluster>
gluster> volume
volume set
set my-dist-volmy-dist-vol auth.allow
192.168.1.*
auth.allow 192.168.1.*
gluster> volume set my-dist-vol auth.reject 10.*
gluster> volume set my-dist-vol auth.reject 10.*
NFS
gluster> volume set my-dist-vol nfs.volume-access read-only
gluster> volume set my-dist-vol nfs.disable on
Other advanced options
gluster> volume set my-dist-vol features.read-only on
gluster> volume set my-dist-vol performance.cache-size 67108864
Volume Top Command
gluster>
gluster> volume
volume set
top my-dist-vol
Brick: server:/export/dir1
my-dist-vol auth.allow
read brick server3:/brick3 192.168.1.*
list-cnt 3
gluster> volume set
==========Read my-dist-vol auth.reject 10.*
file stats========
read filename
call count
116 /clients/client0/~dmtmp/SEED/LARGE.FIL
64 /clients/client0/~dmtmp/SEED/MEDIUM.FIL
54 /clients/client2/~dmtmp/SEED/LARGE.FIL
● Many top commands are available for analysis of
files, directories, and bricks
● Read and write performance test commands available
● Perform active dd tests and measure throughput
Volume Profiling
gluster>
gluster> volume
volume set my-dist-vol
profile my-dist-vol start
gluster> volume profile my-dist-vol info
auth.allow 192.168.1.*
gluster> volume set my-dist-vol
Brick: Test:/export/2 auth.reject 10.*
Cumulative Stats:
Block 1b+ 32b+ 64b+
Size:
Read: 0 0 0
Write: 908 28 8
...
%-latency Avg- Min- Max- calls Fop
latency Latency Latency
___________________________________________________________
4.82 1132.28 21.00 800970.00 4575 WRITE
5.70 156.47 9.00 665085.00 39163 READDIRP
11.35 315.02 9.00 1433947.00 38698 LOOKUP
11.88 1729.34 21.00 2569638.00 7382 FXATTROP
47.35 104235.02 2485.00 7789367.00 488 FSYNC
------------------
Duration : 335
BytesRead : 94505058
BytesWritten : 195571980
Geo-Replication
Setup SSH Keys
# ssh-keygen -f /var/lib/glusterd/geo-replication/secret.pem
# ssh-copy-id -i /var/lib/glusterd/geo-replication/secret.pem repluser@slavehost1
Replicate Via SSH to Remote GlusterFS Volume
gluster> volume geo-replication my-dist-vol repluser@slavehost1::my-dist-repl start
Starting geo-replication session between my-dist-vol & slavehost1:my-dist-repl has been
successful
gluster> volume geo-replication my-dist-vol status
MASTER SLAVE STATUS
--------------------------------------------------------------------------------
my-dist-vol ssh://repluser@slavehost1::my-dist-repl OK
Output of volume info Now Reflects Replication
gluster> volume info my-dist-vol
...
Options Reconfigured:
geo-replication.indexing: on
Use Cases
GlusterFS for SysAdmins
Common Solutions
● Media / Content Distribution Network (CDN)
● Backup / Archive / Disaster Recovery (DR)
● Large Scale File Server
● Home directories
● High Performance Computing (HPC)
● Infrastructure as a Service (IaaS) storage layer
Hadoop – Map Reduce
● Access data within and outside of Hadoop
● No HDFS name node single point of failure / bottleneck
● Seamless replacement for HDFS
● Scales with the massive growth of big data
CIC Electronic Signature Solutions
Hybrid Cloud: Electronic Signature Solutions
● Challenge
● Must leverage economics of the cloud
● Storage performance in the cloud too slow
● Need to meet demanding client SLA’s
● Solution
● Red Hat Storage Software Appliance
● Amazon EC2 and Elastic Block Storage (EBS)
● Reduced time-to- ● Benefits
market for new ● Faster development and delivery of new products
products
● SLA’s met with headroom to spare
● Meeting all client SLAs ● Accelerated cloud migration
● Accelerating move to ● Scale-out for rapid and simple expansion
the cloud ● Data is highly available for 24/7 client access
Pandora Internet Radio
Private Cloud: Media Serving
● Challenge
● Explosive user & title growth
● As many as 12 file formats for each song
● ‘Hot’ content and long tail
● Solution
● Three data centers, each with a six-node GlusterFS
cluster
● Replication for high availability
● 1.2 PB of audio served
per week ● 250+ TB total capacity
● Benefits
● 13 million files
● Easily scale capacity
● Over 50 GB/sec peak ● Centralized management; one administrator to manage
traffic day-to-day operations
● No changes to application
● Higher reliability
Brightcove
Private Cloud: Media Serving
● Challenge
● Explosive customer & title growth
● Massive video in multiple locations
● Costs rising, esp. with HD formats
● Solution
● Complete scale-out based on commodity DAS/JBOD
and GlusterFS
● Replication for high availability
• Over 1 PB currently in ● 1PB total capacity
Gluster ● Benefits
• Separate 4 PB project ● Easily scale capacity
in the works
● Centralized management; one administrator to manage
day-to-day operations
● Higher reliability
● Path to multi-site
Pattern Energy
High Performance Computing for Weather Prediction
●
Challenge
●
Need to deliver rapid advance weather predictions
●
Identify wind and solar abundance in advance
●
More effectively perform preventative maintenance and
repair
●
Solution
●
32 HP compute nodes
• Rapid and advance
weather predictions
●
Red Hat SSA for high throughput and availability
• Maximizing energy
●
20TB+ total capacity
assets ●
Benefits
• Cost savings and ●
Predicts solar and wind patterns 3 to 5 days in advance
avoidance ●
Maximize energy production and repair times
●
Avoid costs of outsourcing weather predictions
●
Solution has paid for itself many times over
Common
Pitfalls
GlusterFS for SysAdmins
Split-Brain Syndrome
● Communication lost between replicated peers
● Clients write separately to multiple copies of a file
● No automatic fix
● May be subjective which copy is right – ALL may be!
● Admin determines the “bad” copy and removes it
● Self-heal will correct the volume
● Trigger a recursive stat to initiate
● Proactive self-healing in GlusterFS 3.3
Quorum Enforcement
● Disallows writes (EROFS) on non-quorum peers
● Significantly reduces files affected by split-brain
● Preferred when data integrity is the priority
● Not preferred when application integrity is the priority
Your Storage Servers are Sacred!
● Don't touch the brick filesystems directly!
● They're Linux servers, but treat them like appliances
● Separate security protocols
● Separate access standards
● Don't let your Jr. Linux admins in!
● A well-meaning sysadmin can quickly break your
system or destroy your data
Do it!
GlusterFS for SysAdmins
Do it!
● Build a test environment in VMs in just minutes!
● Get the bits:
● www.gluster.org has packages available for many
Linux distributions (CentOS, Fedora, RHEL,
Ubuntu)
59
Thank You!
● GlusterFS:
www.gluster.org
● Justin Clift - [email protected]
Slides Available at:
http://www.gluster.org/community/documentation/index.php/Presentations
GlusterFS for SysAdmins
