Understanding the Space Used by ZFS

By Brian Leonard on Sep 28, 2010

Until recently, I've been confused and frustrated by the zfs list output as I try to clear up space on my hard drive.

Take this example using a 1 GB zpool:

bleonard@os200906:~# mkfile 1G /dev/dsk/disk1

bleonard@os200906:~# zpool create tank disk1

bleonard@os200906:~# zpool list tank

NAME   SIZE   USED  AVAIL    CAP  HEALTH  ALTROOT

tank  1016M    73K  1016M     0%  ONLINE  -

Now let's create some files and snapshots:

bleonard@os200906:~# mkfile 100M /tank/file1

bleonard@os200906:~# zfs snapshot tank@snap1

bleonard@os200906:~# mkfile 100M /tank/file2

bleonard@os200906:~# zfs snapshot tank@snap2

bleonard@os200906:~# zfs list -t all -r tank

NAME         USED  AVAIL  REFER  MOUNTPOINT

tank         200M   784M   200M  /tank

tank@snap1    17K      -   100M  -

tank@snap2      0      -   200M  -

The output here looks as I'd expect. I have used 200 MB of disk space, neither of which is used by the snapshots. snap1 refers to 100 MBs of data (file1) and snap2 refers to 200 MBs of data (file1 and file2).

Now let's delete file1 and look at our zfs list output again:

bleonard@os200906:~# rm /tank/file1

bleonard@os200906:~# zpool scrub tank

bleonard@os200906:~# zfs list -t all -r tank

NAME         USED  AVAIL  REFER  MOUNTPOINT

tank         200M   784M   100M  /tank

tank@snap1    17K      -   100M  -

tank@snap2    17K      -   200M  -

Only 1 thing has changed - tank now only refers to 100 MB of data. file1 has been deleted and is only referenced by the snapshots. So why don't the snapshots reflect this in their USED column? You may think we should show 100 MB used by snap1, however, this would be misleading as deleting snap1 has no effect on the data used by the tank file system. Deleting snap1 would only free up 17K of disk space. We'll come back to this test case in a moment.

There is an option to get more detail on the space consumed by the snapshots. Although you can pretty easily deduct from the example above that the snapshots are using 100 MB, by using the zfs space option you can save yourself from doing the math:

bleonard@os200906:~# zfs list -t all -o space -r tank

NAME        AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD

tank         784M   200M      100M    100M              0       154K

tank@snap1      -    17K         -       -              -          -

tank@snap2      -    17K         -       -              -          -

Here we can clearly see that of the 200 MB used by our file system, 100 MB is used by snapshots (file1) and 100 MB is used by the dataset itself (file2). Of course, there are other factors that can affect the total amount used - see the zfs man page for details.

Now, if we were to delete snap1 (we know this is safe, because it's not using any space):

bleonard@os200906:~# zfs destroy tank@snap1

bleonard@os200906:~# zfs list -t all -r tank

NAME         USED  AVAIL  REFER  MOUNTPOINT

tank         200M   784M   100M  /tank

tank@snap2   100M      -   200M  -

We can see that snap2 now shows 100 MBs used. If I were to delete snap2, I would be deleting 100 MB of data (or reclaiming 100 MB of space):

Now let's look at a more realistic example - my home directory where I have Time Slider running:

bleonard@opensolaris:~$ zfs list -t all -r -o space rpool/export/home

NAME                                                       AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD

rpool/export/home                                          25.4G  35.2G     17.9G   17.3G              0          0

rpool/export/home@zfs-auto-snap:monthly-2010-08-03-09:30       -   166M         -       -              -          -

rpool/export/home@zfs-backup-2010-08-12-15:30                  -  5.06M         -       -              -          -

rpool/export/home@zfs-backup-2010-08-12-15:56                  -  5.15M         -       -              -          -

rpool/export/home@zfs-backup-2010-08-31-14:12                  -  54.6M         -       -              -          -

rpool/export/home@zfs-auto-snap:monthly-2010-09-01-00:00       -  53.8M         -       -              -          -

rpool/export/home@zfs-auto-snap:weekly-2010-09-08-00:00        -  95.8M         -       -              -          -

rpool/export/home@zfs-backup-2010-09-09-09:04                  -  53.9M         -       -              -          -

rpool/export/home@zfs-auto-snap:weekly-2010-09-15-00:00        -  2.06G         -       -              -          -

rpool/export/home@zfs-auto-snap:weekly-2010-09-22-00:00        -  89.7M         -       -              -          -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:00      -  18.3M         -       -              -          -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:15      -   293K         -       -              -          -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:30      -   293K         -       -              -          -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:45      -  1.18M         -       -              -          -

My snapshots are consuming almost 18 GBs of space. However, it would appear that I could only reclaim about 2.5 GBs of space by deleting all of my snapshots. In reality, 15.5 GBs of space is referenced by 2 or more snapshots.

I can get a better idea of which snapshots might reclaim the most space by removing the space option so I get the REFER field in the output:

bleonard@opensolaris:~$ zfs list -t all -r rpool/export/home

NAME                                                        USED  AVAIL  REFER  MOUNTPOINT

rpool/export/home                                          35.2G  25.4G  17.3G  /export/home

rpool/export/home@zfs-auto-snap:monthly-2010-08-03-09:30    166M      -  15.5G  -

rpool/export/home@zfs-backup-2010-08-12-15:30              5.06M      -  28.5G  -

rpool/export/home@zfs-backup-2010-08-12-15:56              5.15M      -  28.5G  -

rpool/export/home@zfs-backup-2010-08-31-14:12              54.6M      -  15.5G  -

rpool/export/home@zfs-auto-snap:monthly-2010-09-01-00:00   53.8M      -  15.5G  -

rpool/export/home@zfs-auto-snap:weekly-2010-09-08-00:00    95.8M      -  15.5G  -

rpool/export/home@zfs-backup-2010-09-09-09:04              53.9M      -  17.4G  -

rpool/export/home@zfs-auto-snap:weekly-2010-09-15-00:00    2.06G      -  19.4G  -

rpool/export/home@zfs-auto-snap:weekly-2010-09-22-00:00    89.7M      -  15.5G  -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:15   293K      -  17.3G  -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:30   293K      -  17.3G  -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:45  1.18M      -  17.3G  -

rpool/export/home@zfs-auto-snap:hourly-2010-09-28-12:00        0      -  17.3G  -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-12:00      0      -  17.3G  -

In the above output, I can see that 2 snapshots, taken 26 seconds apart, are referring to 28.5 GBs of disk space. If I were to delete one of those snapshots and check the zfs list output again:

bleonard@opensolaris:~$ pfexec zfs destroy rpool/export/home@zfs-backup-2010-08-12-15:30

bleonard@opensolaris:~$ zfs list -t all -r rpool/export/home

NAME                                                        USED  AVAIL  REFER  MOUNTPOINT

rpool/export/home                                          35.2G  25.4G  17.3G  /export/home

rpool/export/home@zfs-auto-snap:monthly-2010-08-03-09:30    166M      -  15.5G  -

rpool/export/home@zfs-backup-2010-08-12-15:56              12.5G      -  28.5G  -

rpool/export/home@zfs-backup-2010-08-31-14:12              54.6M      -  15.5G  -

rpool/export/home@zfs-auto-snap:monthly-2010-09-01-00:00   53.8M      -  15.5G  -

rpool/export/home@zfs-auto-snap:weekly-2010-09-08-00:00    95.8M      -  15.5G  -

rpool/export/home@zfs-backup-2010-09-09-09:04              53.9M      -  17.4G  -

rpool/export/home@zfs-auto-snap:weekly-2010-09-15-00:00    2.06G      -  19.4G  -

rpool/export/home@zfs-auto-snap:weekly-2010-09-22-00:00    89.7M      -  15.5G  -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:15   293K      -  17.3G  -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:30   293K      -  17.3G  -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-11:45  1.18M      -  17.3G  -

rpool/export/home@zfs-auto-snap:frequent-2010-09-28-12:00   537K      -  17.3G  -

I can now clearly see that the remaining snapshot is using 12.5 GBs of space and deleting this snapshot would reclaim much needed space on my laptop:

bleonard@opensolaris:~$ zpool list rpool

NAME    SIZE   USED  AVAIL    CAP  HEALTH  ALTROOT

rpool   149G   120G  28.5G    80%  ONLINE  -

bleonard@opensolaris:~$ pfexec zfs destroy rpool/export/home@zfs-backup-2010-08-12-15:56

bleonard@opensolaris:~$ zpool list rpool

NAME    SIZE   USED  AVAIL    CAP  HEALTH  ALTROOT

rpool   149G   108G  41.0G    72%  ONLINE  -

And that should be enough to keep Time Slider humming along smoothly and prevent the warning dialog from appearing (lucky you if you haven't seen that yet).

 
 
https://blogs.oracle.com/observatory/entry/understanding_the_space_used_by

Understanding the Space Used by ZFS -- (转)的更多相关文章

  1. linux概念之分区与文件系统

    分区类型 [root@-shiyan dev]# fdisk /dev/sda WARNING: DOS-compatible mode is deprecated. It's strongly re ...

  2. User space 与 Kernel space

    学习 Linux 时,经常可以看到两个词:User space(用户空间)和 Kernel space(内核空间). 简单说,Kernel space 是 Linux 内核的运行空间,User spa ...

  3. Understanding, Operating and Monitoring Apache Kafka

    Apache Kafka is an attractive service because it's conceptually simple and powerful. It's easy to un ...

  4. A Beginner's Guide To Understanding Convolutional Neural Networks(转)

    A Beginner's Guide To Understanding Convolutional Neural Networks Introduction Convolutional neural ...

  5. Understanding CMS GC Logs--转载

    原文地址:https://blogs.oracle.com/poonam/entry/understanding_cms_gc_logs Understanding CMS GC Logs By Po ...

  6. Understanding G1 GC Logs--转载

    原文地址:https://blogs.oracle.com/poonam/entry/understanding_g1_gc_logs Understanding G1 GC Logs By Poon ...

  7. [ZZ] Understanding 3D rendering step by step with 3DMark11 - BeHardware >> Graphics cards

    http://www.behardware.com/art/lire/845/ --> Understanding 3D rendering step by step with 3DMark11 ...

  8. Understanding Virtual Memory

    Understanding Virtual Memory by Norm Murray and Neil Horman Introduction Definitions The Life of a P ...

  9. （转）A Beginner's Guide To Understanding Convolutional Neural Networks Part 2

    Adit Deshpande CS Undergrad at UCLA ('19) Blog About A Beginner's Guide To Understanding Convolution ...

随机推荐

  1. Dubbo学习(四) dubbo的特点，8种通信协议之对比

    一.dubbo的特性 (1) 连通性: 注册中心负责服务地址的注册与查找,相当于目录服务,服务提供者和消费者只在启动时与注册中心交互,注册中心不转发请求,压力较小 监控中心负责统计各服务调用次数,调用 ...

  2. Hibernate Validation，Spring mvc 数据验证框架注解

    1.@NotNull:不能为 Null,但是可以为Empty:用在基本数据类型上. @NotNull(message="{state.notnull.valid}", groups ...

  3. BZOJ5297 CQOI2018社交网络（矩阵树定理）

    板子题. #include<iostream> #include<cstdio> #include<cmath> #include<cstdlib> # ...

  4. P4254 [JSOI2008]Blue Mary开公司

    题面 这道题的意思就是给出若干个一次函数,当\(x=x_0\)时,最大的\(y\)为多少 这种题可以用李超线段树来处理 什么是李超线段树呢? 李超线段树存储的是在区间上方暴露最多的直线标号,为了便于描 ...

  5. P2261 [CQOI2007]余数求和

    我是题面 题意还是很清晰,很容易理解 1e9范围明显不能暴力,除非你能把常数优化到\(\frac1 {10}\),但我实在想象不到用了这么多取模怎么把常数优化下去 我们可以把\(k\%i\)变成\(k ...

  6. Mysql索引机制B+Tree

    1.问题引入 有一个用户表,为了查询的效率,需要基于id去构建索引.构建索引我们需要考虑两个方面的问题,1个是查询的效率,1个是索引数据的存储问题.该表的记录需要支持百万.千万.甚至上亿的数据量,如果 ...

  7. BZOJ 3093: [Fdu校赛2012] A Famous Game

    3093: [Fdu校赛2012] A Famous Game Time Limit: 1 Sec  Memory Limit: 128 MBSubmit: 242  Solved: 129[Subm ...

  8. 主流图片加载框架 ImageLoader、Glide、Picasso、Fresco 对比

    图片缓存库主页: Glidehttps://github.com/bumptech/glide fresco - An Android library for managing images and ...

  9. 微软工程师主讲的SqlServer2005视频教程

    http://www.cnblogs.com/slcfhr/archive/2008/11/24/1340283.html SQL Server 2005盛宴系列之一:SQL Server 2005产 ...

  10. eclipse安装activiti5.18.0工作流插件 以及安装过程中activiti插件出现的问题及解决

    转: eclipse安装activiti5.18.0工作流插件 以及安装过程中activiti插件出现的问题及解决 2017年05月04日 18:44:21 JJ_nan 阅读数:2773   版权声 ...