Category: ZFS

I’ve been meaning to get around to blogging about these features that I
putback a while ago, but have been caught up in a few too many things.
In any case, the following new ZFS features were putback to build 48 of
Nevada, and should be availble in the next Solaris Express

Create Time Properties

An old RFE has been to provide a way to specify properties at create
time. For users, this simplifies admnistration by reducing the number
of commands which need to be run. It also allows some race conditions
to be eliminated. For example, if you want to create a new dataset with
a mountpoint of ‘none’, you first have to create it and the underlying
inherited mountpoint, only to remove it later by invoking ‘zfs set
mountpoint=none’.

From an implementation perspective, this allows us to unify our
implementation of the ‘volsize’ and ‘volblocksize’ properties, and pave
the way for future create-time only properties. Instead of having a
separate ioctl() to create a volume and passing in the two size
parameters, we simply pass them down as create-time options.

The end result is pretty straightforward:

# zfs create -o compression=on tank/home
# zfs create -o mountpoint=/export -o atime=off tank/export

‘canmount’ property

The ‘canmount’ property allows you create a ZFS dataset that serves
solely as a mechanism for inheriting properties. When we first created the
hierarchical dataset model, we had the notion of ‘containers’ –
filesystems with no associated data. Only these datasets could contain
other datasets, and you had to make the decision at create-time.

This turned out to be a bad idea for a number of reasons. It
complicated the CLI, forced the user to make a create-time decision that
could not be changed, and led to confusion when files were accidentally
created on the underlying filesystem. So we made every filesystem able
to have child filesystems, and all seemed well.

However, there is power in having a dataset that exists in the hierarchy
but has no associated filesystem data (or effectively none by preventing
from being mounted). One can do this today by setting the ‘mountpoint’
property to ‘none’. However, this property is inherited by child
datasets, and the administrator cannot leverage the power of inherited
mountpoints. In particular, some users have expressed desire to have
two sets of directories, belonging to different ZFS parents (or even to
UFS filesystems), share the same inherited directory. With the new
‘canmount’ property, this becomes trivial:

# zfs create -o mountpoint=/export -o canmount=off tank/accounting
# zfs create -o mountpoint=/export -o canmount=off tank/engineering
# zfs create tank/accounting/bob
# zfs create tank/engineering/anne

Now, both anne and bob have directories at ‘/export/’, except that
they are inheriting ZFS properties from different datasets in the
hierarchy. The adminsitrator may decide to turn compression on for one
group of people or another, or set a quota to limit the amount of space
consumed by the group. Or simply have a way to view the total amount of
space consumed by each group without resorting to scripted du(1).

User Defined Properties

The last major RFE in this wad added the ability to set arbitrary
properties on ZFS datasets. This provides a way for administrators to
annotate their own filesystems, as well as ISVs to layer intelligent
software without having to modify the ZFS code to introduce a new
property.

A user-defined property name is one which contains a colon (:). This
provides a unique namespace which is guaranteed to not overlap with
native ZFS properties. The emphasis is to use the colon to separate a
module and property name, where ‘module’ should be a reverse DNS name.
For example, a theoretical Sun backup product might do:

# zfs set com.sun.sunbackup:frequency=1hr tank/home

The property value is an arbitrary string, and no additional validation
is done on it. These values are always inherited. A local adminstrator
might do:

# zfs set localhost:backedup=9/19/06 tank/home
# zfs list -o name,localhost:backedup
NAME            LOCALHOST:BACKEDUP
tank            -
tank/home       9/19/06
tank/ws         9/10/06

The hope is that this will serve as a basis for some innovative products
and home grown solutions which interact with ZFS datasets in a
well-defined manner.

More exciting news on the ZFS OpenSolaris front. In addition to the existing ZFS on FUSE/Linux work, we now have a second active port of ZFS, this time for FreeBSD. Pawel Dawidek has been hard at work, and has made astounding progress after just 10 days (!). This is both a testament to his ability as well as the portability of ZFS. As with any port, the hard part comes down to integrating the VFS layer, but Pawel has already made good progress there. The current prototype can already mount fielsystems, create files, and list directory contents. Of course, our code isn’t completely without portability headaches, but thanks to Pawel (and Ricardo on FUSE/Linux), we can take patches and implement the changes upstream to ease future maintenance. You can find the FreeBSD repository Here. If you’re a FreeBSD developer or user, please give Pawel whatever support you can, whether it’s code contributions, testing, or just plain old compliments. We’ll be helping out where we can on the OpenSolaris side.

In related news, Ricard Correia has made significant progress on the FUSE/Linux port. All the management functionality of zfs(1M) and zpool(1M) is there, and he’s working on mounting ZFS filesystems. All in all, it’s an exciting time, and we’re all crossing our fingers that ZFS will follow in the footsteps of its older brother DTrace.

As Jeff mentioned previously, Ricardo Correia has been working on porting ZFS to FUSE/Linux as part of Google SoC. Last week, Ricardo got libzpool and ztest running on Linux, which is a major first step of the project.

The interesting part is the set of changes that he had to make in order to get it working. libzpool was designed to be run from userland and the kernel from the start, so we’ve already done most of the work of separating out the OS-dependent interfaces. The most prolific changes were to satisfy GCC warnings. We do compile ON with gcc, but not using the default options. I’ve since updated the ZFS porting page with info about gcc use in ON, which should make future ports easier. The second most common change was header files that are available in both userland and kernel on Solaris, but nevertheless should be placed in zfs_context.h, concentrating platform-specific knowledge in this one file. Finally, there were some simple changes we could make (such as using pthread_create() instead of thr_create()) to make ports of the tools easier. It would also be helpful to have ports of libnvpair and libavl, much like some have done for libumem, so that developers don’t have to continually port the same libraries over and over.

The next step (getting zfs(1M) and zpool(1M) working) is going to require significantly more changes to our source code. Unlike libzpool, these tools (libzfs in particular) were not designed to be portable. They include a number of Solaris specific interfaces (such as zones and NFS shares) that will be totally different on other platforms. I look forward to seeing Ricardo’s progress to know how this will work out.

It’s been a long time since the last time I wrote a blog entry. I’ve been working heads-down on a new project and haven’t had the time to keep up my regular blogging. Hopefully I’ll be able to keep something going from now on.

Last week the ZFS team put the following back to ON:

PSARC 2006/223 ZFS Hot Spares
PSARC 2006/303 ZFS Clone Promotion
6276916 support for "clone swap"
6288488 du reports misleading size on RAID-Z
6393490 libzfs should be a real library
6397148 fbufs debug code should be removed from buf_hash_insert()
6405966 Hot Spare support in ZFS
6409302 passing a non-root vdev via zpool_create() panics system
6415739 assertion failed: !(zio->io_flags & 0x00040)
6416759 ::dbufs does not find bonus buffers anymore
6417978 double parity RAID-Z a.k.a. RAID6
6424554 full block re-writes need not read data in
6425111 detaching an offline device can result in import confusion

There are a couple of cool features mixed in here. Most importantly, hot spares, clone swap, and double-parity RAID-Z. I’ll focus this entry on hot spares, since I wrote the code for that feature. If you want to see the original ARC case and some of the discussion behind the feature, you should check out the original zfs-discuss thread.

The following features make up hot spare support:

Associating hot spares with pools

Hot spares can be specified when creating a pool or adding devices by using the spare vdev type. For example, you could create a mirrored pool with a single hot spare by doing:

# zpool create test mirror c0t0d0 c0t1d0 spare c0t2d0
# zpool status test
pool: test
state: ONLINE
scrub: none requested
config:
NAME        STATE     READ WRITE CKSUM
test        ONLINE       0     0     0
mirror    ONLINE       0     0     0
c0t0d0  ONLINE       0     0     0
c0t1d0  ONLINE       0     0     0
spares
c0t2d0    AVAIL
errors: No known data errors

Notice that there is one spare, and it currently available for use. Spares can be shared between multiple pools, allowing for a single set of global spares on systems with multiple spares.

Replacing a device with a hot spare

There is now an FMA agent, zfs-retire, which subscribes to vdev failure faults and automatically initiates replacements if there are any hot spares available. But if you want to play around with this yourself (without forcibly faulting drives), you can just use ‘zpool replace’. For example:

# zpool offline test c0t0d0
Bringing device c0t0d0 offline
# zpool replace test c0t0d0 c0t2d0
# zpool status test
pool: test
state: DEGRADED
status: One or more devices has been taken offline by the adminstrator.
Sufficient replicas exist for the pool to continue functioning in a
degraded state.
action: Online the device using 'zpool online' or replace the device with
'zpool replace'.
scrub: resilver completed with 0 errors on Tue Jun  6 08:48:41 2006
config:
NAME          STATE     READ WRITE CKSUM
test          DEGRADED     0     0     0
mirror      DEGRADED     0     0     0
spare     DEGRADED     0     0     0
c0t0d0  OFFLINE      0     0     0
c0t2d0  ONLINE       0     0     0
c0t1d0    ONLINE       0     0     0
spares
c0t2d0      INUSE     currently in use
errors: No known data errors

Note that the offline is optional, but it helps visualize what the pool would look like should and actual device fail. Note that even though the resilver is completed, the ‘spare’ vdev stays in-place (unlike a ‘replacing’ vdev). This is because the replacement is only temporary. Once the original device is replaced, then the spare will be returned to the pool.

Relieving a hot spare

A hot spare can be returned to its previous state by replacing the original faulted drive. For example:

# zpool replace test c0t0d0 c0t3d0
# zpool status test
pool: test
state: DEGRADED
scrub: resilver completed with 0 errors on Tue Jun  6 08:51:49 2006
config:
NAME             STATE     READ WRITE CKSUM
test             DEGRADED     0     0     0
mirror         DEGRADED     0     0     0
spare        DEGRADED     0     0     0
replacing  DEGRADED     0     0     0
c0t0d0   OFFLINE      0     0     0
c0t3d0   ONLINE       0     0     0
c0t2d0     ONLINE       0     0     0
c0t1d0       ONLINE       0     0     0
spares
c0t2d0         INUSE     currently in use
errors: No known data errors
# zpool status test
pool: test
state: ONLINE
scrub: resilver completed with 0 errors on Tue Jun  6 08:51:49 2006
config:
NAME        STATE     READ WRITE CKSUM
test        ONLINE       0     0     0
mirror    ONLINE       0     0     0
c0t3d0  ONLINE       0     0     0
c0t1d0  ONLINE       0     0     0
spares
c0t2d0    AVAIL
errors: No known data errors

The drive is actively being replaced for a short period of time. Once the replacement is completed, the old device is removed, and the hot spare is returned to the list of available spares. If you want a hot spare replacement to become permanent, you can zpool detach the original device, at which point the spare will be removed from the hot spare list of any active pools. You can also zpool detach the spare itself to cancel the hot spare operation.

Removing a spare from a pool

To remove a hot spare from a pool, simply use the zpool remove command. For example:

# zpool remove test c0t2d0
# zpool status
pool: test
state: ONLINE
scrub: resilver completed with 0 errors on Tue Jun  6 08:51:49 2006
config:
NAME        STATE     READ WRITE CKSUM
test        ONLINE       0     0     0
mirror    ONLINE       0     0     0
c0t3d0  ONLINE       0     0     0
c0t1d0  ONLINE       0     0     0
errors: No known data errors

Unfortunately, we don’t yet support removing anything other than hot spares (it’s on our list, we swear). But you can see how hot spares naturally fit into the existing ZFS scheme. Keep in mind that to use hot spares, you will need to upgrade your pools (via ‘zpool upgrade’) to version 3 or later.

Next Steps

Despite the obvious usefulness of this feature, there is one more step that needs to be done for it to be truly useful. This involves phase two of the ZFS/FMA integration. Currently, a drive is only considered faulted if it ‘goes away’ completely (i.e. ldi_open() fails). This covers only subset of known drive failure modes. It’s possible for a drive to continually return errors, and yet be openable. The next phase of ZFS and FMA will introduce a more intelligent diagnosis engine to watch I/O and checksum errors as well as the SMART predictive failure bit in order to proactively offline devices when they are experiencing an abnormal amount of errors, or appear like they are going to fail. With this functionality, ZFS will be able to better respond to failing drives, thereby making hot spare replacement much more valuable.

In the later months of ZFS development, the mechanism used to open and import pools was drastically changed. The reasons behind this change make an interesting case study in complexity management and how a few careful observations can make all the difference.

The original spa_open()

Skipping past some of the very early prototypes, we’ll start with the scheme that Jeff implemented back in 2003. The scheme had a lofty goal:

Be able to open a pool from an arbitrary subset of devices within the pool.

Of course, given limited label space, it was not possible to guarantee this entirely, but we created a scheme which would work in all but the most pessimistic scenarios. Basically, each device was part of a strongly connected graph made up of each toplevel vdev. The graph began with a circle of all the toplevel vdevs. Each device had it’s parent vdev config, the vdev configs of the nearest neighbors, and up to 5 other vdev configs for good measure. When we went to open the pool, we read in this graph and constructed the complete vdev tree which we then used to open the pool.

You can see the original ‘big theory’ comment from the top of vdev_graph.c here.

First signs of problems

The simplest problem to notice was that these vdev configs were stored in a special one-off ASCII representation that required a special parser in the kernel. In a world where we were rapidly transitioning to nvlists, this wasn’t really a good idea.

The first signs of real problems came when I tried to implement the import functionality. For import, we needed the ability to know if a pool could be imported without actually opening it. This had existed in a rather disgusting form previously by linking to libzpool (which is the userland port of the SPA and DMU code), and hijacking the versions of these functions to construct a vdev tree in userland. In the new userland model, it wasn’t acceptable to use libzpool in this manner. So, I had to construct spa_open_by_dev() that parsed the configuration into a vdev tree and tried to open the vdevs, but not actually load the pool metadata.

This required a fair amount of hackery, but it was nowhere near as bad as when I had to make the resulting system tolerant of all faults. For both ‘zpool import’ and ‘zpool status’, it wasn’t enough just to know that a pool couldn’t be opened. We needed to know why, and exactly which devices were at fault. While the original scheme worked well when a single device was missing in a toplevel vdev, it failed miserably when an entire toplevel vdev was missing. In particular, it relied on the fact that it had at least a complete ring of toplevel vdevs to work with. For example, you were missing a single device in an unmirrored pool, there was no way to know what device was missing because the entire graph parsing algorithm would break down. So, I went in and hacked on the code to understand multiple “versions” of a vdev config. If we had two neighbors referring to a missing toplevel vdev, we could surmise its config even though all its devices were absent.

At this point, things were already getting brittle. The code was enormously complex, and hard to change in any well defined way. On top of that, things got even worse when our test guys started getting really creative. In particular, if you disconnected a device and then exported a pool, or created a new pool over parts of an exported pool, things would get ugly fast. The labels on the disks would be technically valid, but semantically invalid. I had to make even more changes to the algorithms to accomodate all these edge cases. Eventually, it got to the point where every change I made was prefixed by /* XXX kludge */. Jeff and I decided something needed to be done.

On top of all this, the system still had a single point of failure. Since we didn’t want to scan every attached device on boot, we kept an /etc/system file around that described the first device in the pool as a ‘hint’ for where to get started. If this file was corrupt, or that particular toplevel vdev was not available, the pool configuration could not be determined.

vdev graph 2.0

At this point we had a complex, brittle, and functionally insufficient system for discovering pools. As Jeff, Bill, and I talked about this problem for a while, we made two key observations:

  • The kernel didn’t have parse the graph. We already had the case where we were dependent on a cached file for opening our pool, so why not keep the whole configuration there? The kernel can do (relatively) atomic updates to this file on configuration changes, and then open the resulting vdev tree without having to construct it based on on-disk data.
  • During import, we already need to check all devices in the system. We don’t have to worry about ‘discovering’ other toplevel vdevs, because we know that we will, by definition, look at the device during the discovery phase.

With these two observations under our belt, we knew what we had to do. For both open and import, the SPA would know only how to take a fully constructed config and parse it into a working vdev tree. Whether that config came from the new /etc/zfs/zpool.cache file, or whether it was constructed during a ‘zpool import’ scan, it didn’t matter. The code would be exactly the same. And best of all, no complicated discovery phase – the config was taken at face value1. And the config was simply an nvlist. No more special parsing of custom ‘vdev spec’ strings.

So how does import work?

The config cache was all well and good, but how would import work? And how would it fare in the face of total device failure? You can find all of the logic for ‘zpool import’ in libzfs_import.c.

Each device keeps the complete config for its toplevel vdev. No neighbors, no graph, no nothing. During pool discovery, we keep track of all toplevel vdevs for a given pool that we find during a scan. Using some simple heuristics, we construct the ‘best’ version of the pool’s config, and even go through and update the path and devid information based on the unique GUID for each device. Once that’s done, we have the same nvlist we would have as if we had read it from zpool.cache. The kernel happily goes off and tries to open the vdevs (if this is just a scan) or open it for real (if we’re doing the import).

So what about missing toplevel vdevs? In the online (cached) state, we’ll have the complete config and tell you which device is missing. For the import case, we’ll be a little worse off because we’ll never see any vdevs indicating that there is another toplevel vdev in the pool. The most important thing is that we’re able to detect this case and error out properly. To do this, we have a ‘vdev guid sum’ stored in the uberblock that indicates the sum of all the vdev guids for every device in the config. If this doesn’t match, we know that we have missed a toplevel vdev somewhere. Unfortunately, we can’t tell you what device it is. In the future, we hope to improve this by adding the concept of ‘neighbor lists’ – arbitrary lists of devices without entire vdev configs. Unlike the previous incarnation of vdev graph, these will be purely suggestions, and not actually be required for correctness. There will, of course, be cases where we can never provide you enough information about all your neighbors, such as plugging in a single disk from a thousand disk unreplicated pool.

Conclusions

So what did we learn from this? The main thing is that phrasing the problem slightly differently can cause you to overdesign a system beyond the point of maintainability. As Bryan is fond of saying, one of the most difficult parts of solving hard problems is not just working within a highly constrained environment, but identifying which constraints aren’t needed at all. By realizing that opening a pool was a very different operation than discovering a pool, we were able to redesign our interfaces into a much more robust and maintainable state, with more features to boot.

It’s no surprise that there were more than a few putbacks like “SPA 2.0”, “DMU 2.0”, or “ZIL 3.0”. Sometimes you just need to take a hammer to the foundation to incorporate all that you’ve learned from years of using the previous version.


1 Actually, Jeff made this a little more robust by adding the config as part of the MOS (Meta objset), which is stored transactionally with the rest of the pool data. So even if we added two devices, but the /etc cache file didn’t get updated correctly, we’ll still be able to open the MOS config and realize that there are two new devices to be had.

In this post I’ll describe the interactions between ZFS and FMA (Fault Management Architecture). I’ll cover the support that’s present today, as well as what we’re working on and where we’re headed.

ZFS Today (phase zero)

The FMA support in ZFS today is what we like to call “phase zero”. It’s basically the minimal amount of integration needed in order to leverage (arguably) the most useful feature in FMA: knowledge articles. One of the key FMA concepts is to present faults in a readable, consistent manner across all subsystems. Error messages are human readable, contain precise descriptions of what’s wrong and how to fix it, and point the user to a website that can be updated more frequently with the latest information.

ZFS makes use of this strategy in the ‘zpool status’ command:

$ zpool status
pool: tank
state: ONLINE
status: One or more devices has experienced an unrecoverable error.  An
attempt was made to correct the error.  Applications are unaffected.
action: Determine if the device needs to be replaced, and clear the errors
using 'zpool online' or replace the device with 'zpool replace'.
see: http://www.sun.com/msg/ZFS-8000-9P
scrub: none requested
config:
NAME        STATE     READ WRITE CKSUM
tank        ONLINE       0     0     0
mirror    ONLINE       0     0     0
c1d0s0  ONLINE       0     0     3
c0d0s0  ONLINE       0     0     0

In this example, one of our disks has experienced multiple checksum errors (because I dd(1)ed over most of the disk), but it was automatically corrected thanks to the mirrored configuration. The error message described exactly what has happened (we tried to self-heal the data from the other side of the mirror) and the impact (none – applications are unaffected). It also directs the user to the appropriate repair procedure, which is either to clear the errors (if they are not indicative of hardware fault) or replace the device.

It’s worth noting here the ambiguity of the fault. We don’t actually know if the errors are due to bad hardware, transient phenomena, or administrator error. More on this later.

ZFS tomorrow (phase one)

If we look at the implementation of ‘zpool status’, we see that there is no actual interaction with the FMA subsystem apart from the link to the knowledge article. There are no generated events or faults, and no diagnosis engine or agent to subscribe to the events. The implementation is entirely static, and contained within libzfs_status.c.

This is obviously not an ideal solution. It doesn’t give the administrator any notification of errors as they occur, nor does it allow them to leverage other pieces of the FMA framework (such as upcoming SNMP trap support). This is going to be addressed in the near term by the first “real” phase of FMA integration. The goal of this phase, in addition to a number of other fault capabilities under the hood, is the following:

  • Event Generation – I/O errors and vdev transitions will result in true FMA ereports. These will be generated by the SPA and fed through the FMA framework for further analysis.
  • Simple Diagnosis Engine – An extremely dumb diagnosis engine will be provided to consume these ereports. It will not perform any sort of predictive analysis, but will be able to keep track of whether these errors have been seen before, and pass them off to the appropriate agent.
  • Syslog agent – The results from the DE will be fed to an agent that simply forwards the faults to syslog for the administrator’s benefit. This will give the same messages as seen in ‘zpool status’ (with slightly less information) synchronous with a fault event. Future work to generate SNMP traps will allow the administrator to be email, paged, or implement a poor man’s hot spare system.

ZFS in the future (phase next)

Where we go from here is rather an open road. The careful observer will notice that ZFS never makes any determination that a device is faulted due to errors. If the device fails to reopen after an I/O error, we will mark it as faulted, but this only catches cases where a device has gotten completely out of whack. Even if your device is experiencing a hundred uncorrectable I/O errors per second, ZFS will continue along its merry way, notifying the administrator but otherwise doing nothing. This is not because we don’t want to take the device offline; it’s just that getting the behavior right is hard.

What we’d like to see is some kind of predictive analysis of the error rates, in an attempt to determine if a device is truly damaged, or whether it was just a random event. The diagnosis engines provided by FMA are designed for exactly this, though the hard part is determining the algorithms for making this distinction. ZFS is both a consumer of FMA faults (in order to take proactive action) as well as a producer of ereports (detecting checksum errors). To be done right, we need to harden all of our I/O drivers to generate proper FMA ereports, implement a generic I/O retire mechanism, and link it in with all the additional data from ZFS. We also want to gather SMART data from the drives to notice correctible errors fixed by the firmware, as well doing experiments to determine the failure rates and pathologies of common storage drives.

As you can tell, this is not easy. I/O fault diagnosis is much more complicated than CPU and memory diagnosis. There are simply more components, as well as more changes for administrative error. But between FMA and ZFS, we have laid the groundwork for a truly fault-tolerant system capable of predictive self healing and automated recovery. Once we get past the initial phase above, we’ll start to think about this in more detail, and make our ideas public as well.

Recent Posts

April 21, 2013
February 28, 2013
August 14, 2012
July 28, 2012

Archives