github.com/boynux/docker@v1.11.0-rc4/docs/userguide/storagedriver/zfs-driver.md (about) 1 <!--[metadata]> 2 +++ 3 title = "ZFS storage in practice" 4 description = "Learn how to optimize your use of ZFS driver." 5 keywords = ["container, storage, driver, ZFS "] 6 [menu.main] 7 parent = "engine_driver" 8 +++ 9 <![end-metadata]--> 10 11 # Docker and ZFS in practice 12 13 ZFS is a next generation filesystem that supports many advanced storage 14 technologies such as volume management, snapshots, checksumming, compression 15 and deduplication, replication and more. 16 17 It was created by Sun Microsystems (now Oracle Corporation) and is open sourced 18 under the CDDL license. Due to licensing incompatibilities between the CDDL 19 and GPL, ZFS cannot be shipped as part of the mainline Linux kernel. However, 20 the ZFS On Linux (ZoL) project provides an out-of-tree kernel module and 21 userspace tools which can be installed separately. 22 23 The ZFS on Linux (ZoL) port is healthy and maturing. However, at this point in 24 time it is not recommended to use the `zfs` Docker storage driver for 25 production use unless you have substantial experience with ZFS on Linux. 26 27 > **Note:** There is also a FUSE implementation of ZFS on the Linux platform. 28 > This should work with Docker but is not recommended. The native ZFS driver 29 > (ZoL) is more tested, more performant, and is more widely used. The remainder 30 > of this document will relate to the native ZoL port. 31 32 33 ## Image layering and sharing with ZFS 34 35 The Docker `zfs` storage driver makes extensive use of three ZFS datasets: 36 37 - filesystems 38 - snapshots 39 - clones 40 41 ZFS filesystems are thinly provisioned and have space allocated to them from a 42 ZFS pool (zpool) via allocate on demand operations. Snapshots and clones are 43 space-efficient point-in-time copies of ZFS filesystems. Snapshots are 44 read-only. Clones are read-write. Clones can only be created from snapshots. 45 This simple relationship is shown in the diagram below. 46 47  48 49 The solid line in the diagram shows the process flow for creating a clone. Step 50 1 creates a snapshot of the filesystem, and step two creates the clone from 51 the snapshot. The dashed line shows the relationship between the clone and the 52 filesystem, via the snapshot. All three ZFS datasets draw space form the same 53 underlying zpool. 54 55 On Docker hosts using the `zfs` storage driver, the base layer of an image is a 56 ZFS filesystem. Each child layer is a ZFS clone based on a ZFS snapshot of the 57 layer below it. A container is a ZFS clone based on a ZFS Snapshot of the top 58 layer of the image it's created from. All ZFS datasets draw their space from a 59 common zpool. The diagram below shows how this is put together with a running 60 container based on a two-layer image. 61 62  63 64 The following process explains how images are layered and containers created. 65 The process is based on the diagram above. 66 67 1. The base layer of the image exists on the Docker host as a ZFS filesystem. 68 69 This filesystem consumes space from the zpool used to create the Docker 70 host's local storage area at `/var/lib/docker`. 71 72 2. Additional image layers are clones of the dataset hosting the image layer 73 directly below it. 74 75 In the diagram, "Layer 1" is added by making a ZFS snapshot of the base 76 layer and then creating a clone from that snapshot. The clone is writable and 77 consumes space on-demand from the zpool. The snapshot is read-only, maintaining 78 the base layer as an immutable object. 79 80 3. When the container is launched, a read-write layer is added above the image. 81 82 In the diagram above, the container's read-write layer is created by making 83 a snapshot of the top layer of the image (Layer 1) and creating a clone from 84 that snapshot. 85 86 As changes are made to the container, space is allocated to it from the 87 zpool via allocate-on-demand operations. By default, ZFS will allocate space in 88 blocks of 128K. 89 90 This process of creating child layers and containers from *read-only* snapshots 91 allows images to be maintained as immutable objects. 92 93 ## Container reads and writes with ZFS 94 95 Container reads with the `zfs` storage driver are very simple. A newly launched 96 container is based on a ZFS clone. This clone initially shares all of its data 97 with the dataset it was created from. This means that read operations with the 98 `zfs` storage driver are fast – even if the data being read was note 99 copied into the container yet. This sharing of data blocks is shown in the 100 diagram below. 101 102  103 104 Writing new data to a container is accomplished via an allocate-on-demand 105 operation. Every time a new area of the container needs writing to, a new block 106 is allocated from the zpool. This means that containers consume additional 107 space as new data is written to them. New space is allocated to the container 108 (ZFS Clone) from the underlying zpool. 109 110 Updating *existing data* in a container is accomplished by allocating new 111 blocks to the containers clone and storing the changed data in those new 112 blocks. The original blocks are unchanged, allowing the underlying image 113 dataset to remain immutable. This is the same as writing to a normal ZFS 114 filesystem and is an implementation of copy-on-write semantics. 115 116 ## Configure Docker with the ZFS storage driver 117 118 The `zfs` storage driver is only supported on a Docker host where 119 `/var/lib/docker` is mounted as a ZFS filesystem. This section shows you how to 120 install and configure native ZFS on Linux (ZoL) on an Ubuntu 14.04 system. 121 122 ### Prerequisites 123 124 If you have already used the Docker daemon on your Docker host and have images 125 you want to keep, `push` them Docker Hub or your private Docker Trusted 126 Registry before attempting this procedure. 127 128 Stop the Docker daemon. Then, ensure that you have a spare block device at 129 `/dev/xvdb`. The device identifier may be be different in your environment and 130 you should substitute your own values throughout the procedure. 131 132 ### Install Zfs on Ubuntu 14.04 LTS 133 134 1. If it is running, stop the Docker `daemon`. 135 136 1. Install the `software-properties-common` package. 137 138 This is required for the `add-apt-repository` command. 139 140 $ sudo apt-get install -y software-properties-common 141 Reading package lists... Done 142 Building dependency tree 143 <output truncated> 144 145 2. Add the `zfs-native` package archive. 146 147 $ sudo add-apt-repository ppa:zfs-native/stable 148 The native ZFS filesystem for Linux. Install the ubuntu-zfs package. 149 <output truncated> 150 gpg: key F6B0FC61: public key "Launchpad PPA for Native ZFS for Linux" imported 151 gpg: Total number processed: 1 152 gpg: imported: 1 (RSA: 1) 153 OK 154 155 3. Get the latest package lists for all registered repositories and package 156 archives. 157 158 $ sudo apt-get update 159 Ign http://us-west-2.ec2.archive.ubuntu.com trusty InRelease 160 Get:1 http://us-west-2.ec2.archive.ubuntu.com trusty-updates InRelease [64.4 kB] 161 <output truncated> 162 Fetched 10.3 MB in 4s (2,370 kB/s) 163 Reading package lists... Done 164 165 4. Install the `ubuntu-zfs` package. 166 167 $ sudo apt-get install -y ubuntu-zfs 168 Reading package lists... Done 169 Building dependency tree 170 <output truncated> 171 172 5. Load the `zfs` module. 173 174 $ sudo modprobe zfs 175 176 6. Verify that it loaded correctly. 177 178 $ lsmod | grep zfs 179 zfs 2768247 0 180 zunicode 331170 1 zfs 181 zcommon 55411 1 zfs 182 znvpair 89086 2 zfs,zcommon 183 spl 96378 3 zfs,zcommon,znvpair 184 zavl 15236 1 zfs 185 186 ## Configure ZFS for Docker 187 188 Once ZFS is installed and loaded, you're ready to configure ZFS for Docker. 189 190 191 1. Create a new `zpool`. 192 193 $ sudo zpool create -f zpool-docker /dev/xvdb 194 195 The command creates the `zpool` and gives it the name "zpool-docker". The name is arbitrary. 196 197 2. Check that the `zpool` exists. 198 199 $ sudo zfs list 200 NAME USED AVAIL REFER MOUNTPOINT 201 zpool-docker 55K 3.84G 19K /zpool-docker 202 203 3. Create and mount a new ZFS filesystem to `/var/lib/docker`. 204 205 $ sudo zfs create -o mountpoint=/var/lib/docker zpool-docker/docker 206 207 4. Check that the previous step worked. 208 209 $ sudo zfs list -t all 210 NAME USED AVAIL REFER MOUNTPOINT 211 zpool-docker 93.5K 3.84G 19K /zpool-docker 212 zpool-docker/docker 19K 3.84G 19K /var/lib/docker 213 214 Now that you have a ZFS filesystem mounted to `/var/lib/docker`, the daemon 215 should automatically load with the `zfs` storage driver. 216 217 5. Start the Docker daemon. 218 219 $ sudo service docker start 220 docker start/running, process 2315 221 222 The procedure for starting the Docker daemon may differ depending on the 223 Linux distribution you are using. It is possible to force the Docker daemon 224 to start with the `zfs` storage driver by passing the 225 `--storage-driver=zfs`flag to the `docker daemon` command, or to the 226 `DOCKER_OPTS` line in the Docker config file. 227 228 6. Verify that the daemon is using the `zfs` storage driver. 229 230 $ sudo docker info 231 Containers: 0 232 Images: 0 233 Storage Driver: zfs 234 Zpool: zpool-docker 235 Zpool Health: ONLINE 236 Parent Dataset: zpool-docker/docker 237 Space Used By Parent: 27648 238 Space Available: 4128139776 239 Parent Quota: no 240 Compression: off 241 Execution Driver: native-0.2 242 [...] 243 244 The output of the command above shows that the Docker daemon is using the 245 `zfs` storage driver and that the parent dataset is the 246 `zpool-docker/docker` filesystem created earlier. 247 248 Your Docker host is now using ZFS to store to manage its images and containers. 249 250 ## ZFS and Docker performance 251 252 There are several factors that influence the performance of Docker using the 253 `zfs` storage driver. 254 255 - **Memory**. Memory has a major impact on ZFS performance. This goes back to 256 the fact that ZFS was originally designed for use on big Sun Solaris servers 257 with large amounts of memory. Keep this in mind when sizing your Docker hosts. 258 259 - **ZFS Features**. Using ZFS features, such as deduplication, can 260 significantly increase the amount of memory ZFS uses. For memory consumption 261 and performance reasons it is recommended to turn off ZFS deduplication. 262 However, deduplication at other layers in the stack (such as SAN or NAS arrays) 263 can still be used as these do not impact ZFS memory usage and performance. If 264 using SAN, NAS or other hardware RAID technologies you should continue to 265 follow existing best practices for using them with ZFS. 266 267 - **ZFS Caching**. ZFS caches disk blocks in a memory structure called the 268 adaptive replacement cache (ARC). The *Single Copy ARC* feature of ZFS allows a 269 single cached copy of a block to be shared by multiple clones of a filesystem. 270 This means that multiple running containers can share a single copy of cached 271 block. This means that ZFS is a good option for PaaS and other high density use 272 cases. 273 274 - **Fragmentation**. Fragmentation is a natural byproduct of copy-on-write 275 filesystems like ZFS. However, ZFS writes in 128K blocks and allocates *slabs* 276 (multiple 128K blocks) to CoW operations in an attempt to reduce fragmentation. 277 The ZFS intent log (ZIL) and the coalescing of writes (delayed writes) also 278 help to reduce fragmentation. 279 280 - **Use the native ZFS driver for Linux**. Although the Docker `zfs` storage 281 driver supports the ZFS FUSE implementation, it is not recommended when high 282 performance is required. The native ZFS on Linux driver tends to perform better 283 than the FUSE implementation. 284 285 The following generic performance best practices also apply to ZFS. 286 287 - **Use of SSD**. For best performance it is always a good idea to use fast 288 storage media such as solid state devices (SSD). However, if you only have a 289 limited amount of SSD storage available it is recommended to place the ZIL on 290 SSD. 291 292 - **Use Data Volumes**. Data volumes provide the best and most predictable 293 performance. This is because they bypass the storage driver and do not incur 294 any of the potential overheads introduced by thin provisioning and 295 copy-on-write. For this reason, you should place heavy write workloads on data 296 volumes.