> - while you can enable on a filesystem level it operates on pool
> level. once activated you cannot disable beside a pool destroy.
As far as I can tell from having looked at the code in the past,
dedup is effectively inactive as soon as all data written while
dedup was enabled has been
> Yes I saw that too, but I'm not sure why 12GB is being taken up by
> rpool/dump. And there's really not that much used by BEs. Is there a
> reason that dumps has to be so large or should I allocate more space
> to rpool? Thanks!
If I remember correctly, by default rpool/dump is sized during
> Which does open the question: how many people need LTS? (And what does
> LTS mean for them - in terms of how long support would be needed, and
> what level of support/backports they expect.) Perhaps Chris could chip
> in here, but I know that with my $DAYJOB hat on the idea of dropping
>
The OmniOS CE release announcement here on the mailing list covered
the broad plans for the (non-Bloody) release schedule:
The intention is for new stable releases to continue to come
out every 26 weeks. Interim, "weekly" updates to stable follow a
fixed schedule denoted
> Hi,
> I am using the dtrace script nfsv3rwsnoop.d to find file that are
> accessed from my OmniOS r151020 filer and some file names are listed as
> unknown :-(
> I guess they are files that have been open for a long time and have
> dropped out of some data structure.
> Is there any way to
> Im looking for some general dtrace scripts for debugging ZFS on OmniOS
> (like updated dtrace toolkit)..didnt want to reinvent the wheel if
> some folks are willing to share. Also willing to purchase if needed.
We have a collection of local DTrace scripts at:
For the interest of bystanders, here's the resolution to our puzzle.
I wrote:
> I wrote, about our mysteriously disappearing snapshots:
> > [...] For example, is there some way where snapshots can be removed
> > without that being logged in 'zpool history'?
>
> The answer to my question
> > On Jan 9, 2017, at 11:13 AM, Chris Siebenmann <c...@cs.toronto.edu> wrote:
> >
> > The answer to my question turns out to be 'yes'. If you do:
> > rmdir /.zfs/snapshot/
> >
> > and ZFS accepts this, there is no entry made in 'zpool history'; the
I wrote, about our mysteriously disappearing snapshots:
> [...] For example, is there some way where snapshots can be removed
> without that being logged in 'zpool history'?
The answer to my question turns out to be 'yes'. If you do:
rmdir /.zfs/snapshot/
and ZFS accepts this, there
> I wonder if the scrub has something to do with it? Are you in the
> middle of the scrub when noticing missing snapshots? Do they reappear
> after the scrub is done? The scrub seems like it's relevant. I take
> it you're doing a cron-driven scrub? If so, with what frequency?
We scrub
I wrote, quoting first 'zpool history' and then 'zfs list' output:
[...]
> 2017-01-06.15:10:01 zfs snapshot fs0-admin-02/h/105@Fri-15
> 2017-01-06.16:10:01 zfs snapshot fs0-admin-02/h/105@Fri-16
> 2017-01-06.16:45:55 zfs snapshot fs0-admin-02/h/105@Fri-16
>
We have an automated system for making regular (roughly hourly)
snapshots of some especially important filesystems where we want
fast restores. This has been running smoothly for some time and
without problems. However, starting this week we have twice gone
to do from-snapshot restores on one of
We recently had a server reboot due to the ZFS vdev_deadman/spa_deadman
timeout timer activating and panicing the system. If you haven't heard
of this timer before, that's not surprising; triggering it requires an
IO to a vdev to take more than 1000 seconds (by default; it's controlled
by the
> I got a link to this commit from the Delphix illumos repo a while back:
>
> https://github.com/openzfs/openzfs/pull/186/
>
> I was curious if NFS-using people in the audience here would like to
> see this one Just Land (TM) in illumos-omnios or not?
I think that modernizing these NFS
> Does anybody uses ZFS over iSCSI?
>
> There is problem with reboots as iscsi-initiator service does not
> taking care of ZFS while shutting down. It leads to zpool goes into
> UNAVAIL state and then first sync() issued gets blocked with following
> stack:
We have a significant ZFS-over-iSCSI
[About ZFS scrub tunables:]
> Interesting read - and it surely works. If you set the tunable before
> you start the scrub you can immediately see the thoughput being much
> higher than with the standard setting. [...]
It's perhaps worth noting here that the scrub rate shown in 'zpool
status' is
> So my quick question to you all: for r151018 (and backporting), do you
> prefer a nicer-looking installer that craps out if you select Africa,
> Europe, or Asia? Or do you prefer a sketchier-looking one that works
> for all timezones out of the box? I'm cutting the last 017 bloody
> with this,
I have a relatively complicated chunk of dtrace code that reads kernel
data structures and chases pointers through them. Some of the time it
spits out 'invalid address' errors during execution, for example:
dtrace: error on enabled probe ID 8 (ID 75313:
fbt:nfssrv:nfs3_fhtovp:return):
> > The sd.conf whitelist also requires a reboot to activate if you need
> > to add a new entry, as far as I know.
> >
> > (Nor do I know what happens if you have some 512n disks and
> > some 512e disks, both correctly recognized and in different
> > pools, and now you need to
> > This implicitly assumes that the only reason to set ashift=12 is
> > if you are currently using one or more drives that require it. I
> > strongly disagree with this view. Since ZFS cannot currently replace
> > a 512n drive with a 512e one, I feel [...]
>
> *In theory* this replacement
> > Adding the ashift argument to zpool was discussed every few years
> > and so far was always deemed not enterprisey enough for the Solaris
> > heritage, so the setup to tweak sd driver reports and properly rely
> > on that layer was pushed instead.
>
> The issue is that once a drive model lies,
> > > Motherboard - Supermicro X10DRi family
> > > CPUs - E5-2620 v3
> > > HBA - LSI 9300-8i
> > > Network - Intel i350 or 540, depending on precise motherboard variant
> > > Disks (SSD) - Intel S3510 (boot), S3610 (application + data)
> >
> > The drives are SATA and there are quite a few SATA
> I have to replace a HDD disk in a mirrored vdev which is using
> ashift=9. The problem is that all new HDD's is 4K disks so I wonder
> if anyone has made some performance measurements of using a 4K disk
> with ashift=9 in contrast to the obvious ashift=12?
>
> As I understand it you cannot
It's great that Sun SSH can now be more or less transparently swapped
for OpenSSH. As part of that both of them provide the SMF service
svc:/network/ssh:default. However, there is a difference in the SMF
properties that each of them uses for boot time ordering: Sun SSH uses
fs-local/*, and
> Technically, I believe OmniOS "owns" /opt/omni. Feel free to define
> other subdirs as datasets hosted elsewhere (out of tree of rootfs
> children). My own admin-scripts go under /opt but these are okay as
> part of rootfs/BE so I don't dataset them ;)
OmniOS also puts stuff in eg
> My first try at an update failed to boot properly. Being a control
> freak, I had made /opt its own zfs filesystem. Previously it must
> have been part of the boot environment. The contents of /opt was not
> as expected and so the fs.local service failed.
We ran into this in our ongoing
> Sure, generally speaking. In this particular context I believe users
> should ship their own if they want to deploy a mail server, but
> all nodes should be able to deliver mail locally. It would also be
> great if the default install lended itself to mail submission (eg.
> a satellite mailer
We have a situation where we would like to be able to install new
r151014 machines with something other than the current r151014 kernel.
(In the extreme case we'd like to be able to specify the exact package
version for all packages, but kernels are the most important for us.)
I *think* that
> I completely concur with Richard on this. Let me give an a real example
> that emphases this point as it's a critical design decision.
[...]
> Now I only run one hot spare per pool. Most of my pools are raidz2 or
> raidz3. This way any event like this can not take out more than one
> disk and
I wrote:
We have one ZFS-based NFS fileserver that persistently runs at a very
high level of non-ARC kernel memory usage that never seems to shrink.
On a 128 GB machine, mdb's ::memstat reports 95% memory usage by just
'Kernel' while the ZFS ARC is only at about 21 GB (as reported by
'kstat
It turns out that the explanation for this is relatively simple, as
is the work around. Put simply: the OmniOS kernel does not actually
free up these deallocated cache objects until the system is put
under relatively strong memory pressure. Crucially, *the ZFS ARC
does not create this
We have one ZFS-based NFS fileserver that persistently runs at a very
high level of non-ARC kernel memory usage that never seems to shrink.
On a 128 GB machine, mdb's ::memstat reports 95% memory usage by just
'Kernel' while the ZFS ARC is only at about 21 GB (as reported by
'kstat -m') although
According to the OmniOS source repository's changelogs, there's a
fix in the r151014 branch that we're quite interested in:
3783 Flow control is needed in rpcmod when the NFS server is
unable to keep up with the network
As far as I know this fix is not in the initial release
On May 4, 2015, at 9:03 PM, Dan McDonald dan...@omniti.com wrote:
I swear I've seen someone try to address this before. Maybe it's from =
my Nexenta days. I will be querying the illumos developer's list (as I =
suspect this affects the other distros as well if they haven't fixed it =
in
The hard part will be testing this. I'm not sure I have the HW
in-house to do it. I may need illumos community help.
Since we have a test environment where we can reproduce this and a
high interest in seeing it fixed, we can test new kernel packages
and so on.
(If given specific
We now have a reproducable setup with OmniOS r151014 where an OmniOS
NFS fileserver will experience memory exhaustion and then hang in the
kernel if it receives sustained NFS write traffic from multiple clients
at a rate faster than its local disks can sustain. The machine will run
okay for a
On Apr 29, 2015, at 3:21 PM, Chris Siebenmann c...@cs.toronto.edu wrote:
We have a filesystem/dataset with no snapshots,
You're sure about no snapshots? zfs list -t snapshot has surprised
me once or twice in the past. :-/
Completely sure. 'zfs list -t snapshot' has nothing and all
We have a filesystem/dataset with no snapshots, no subordinate
filesystems, nothing complicated (and no compression), that has a
drastic difference in space used between what df/zfs list/etc report at
the ZFS level and what du reports at the filesystem level. ZFS says
NAME PROPERTY
From: Chris Siebenmann
[...] ZFS level; as you can see, it's added 22 GB of ZFS usage in
less than a month while losing 5 GB at the user level.
There was a thread late last year on the developer list that described
similar symptoms:
http://lists.open-zfs.org/pipermail/developer/2014
We now have a reproducable situation where high NFS load can cause our
specific fileserver configuration to lock up with what looks like a
memory exhaustion deadlock. So far, attempts to get a crash dump haven't
worked (although they may someday). Without a crash dump, what system
stats and so on
Oddly, removing the SFP modules makes no difference. Replacing the
modules - still doesn't show up. I really do expect to see the
interfaces in dladm show-phys now.. also add_drv still fails to
attach, which is odd. I feel like this has something to do with fault
management not clearing
Out of curiosity: I suspect that plenty of people are gathering basic
system activity stats for their OmniOS systems and pushing them into
modern metrics systems such as graphite (to pick perhaps the most well
known package for this). For those that are doing this, what is your
preferred
History lesson: until people could afford to purchase more than one
disk and before Sun invented the diskless workstation (with shared
/usr), everything was under /.
As Richard knows but other people may not, this is ahistorical on
Unix. From almost the beginning[*] Unix had a split between
Short story is that /opt is part of a namespace managed by the Solaris
packaging and as such is part of a BE fs tree. If you have privately
managed packages under certain subdirs, turn those sub-dirs into
separate datasets instead.
If this is the case for OmniOS, I believe that it should be
A better way is to IPS package everything properly, and add proper
metadata to your packages, so that those packages that should go
into a new BE do ask for one in their manifest. That way, there
is no distinction between any system package living in /usr (or
wherever) and your package
We're running into a situation with one of our NFS ZFS fileservers[*]
where we're wondering if we have enough NFS server threads to handle
our load. Per 'sharectl get nfs', we have 'servers=512' configured,
but we're not sure we know how to check how many are actually in use
and active at any
After installation and configuration, I observed all kinds of bad behavior
in the network traffic between the hosts and the server. All of this bad
behavior is traced to the ixgbe driver on the storage server. Without going
into the full troubleshooting process, here are my takeaways:
[...]
If you have a ZFS pool with mirror vdevs and you look at 'zfs list'
versus 'zpool list', you can see that the available space is somewhat
different between the two. This is a known issue and comes about because
the ZFS code reserves some amount of space that can't be consumed in
normal use and
We have found a reproducable iSCSI initiator issue in our environment.
The short form version of the hang is that if there is a network
interruption between one of our OmniOS machines and an iSCSI target and
you run 'iscsiadm list target -S' in a roughly 30-second window after
the interruption
We have a situation where one of our OmniOS NFS fileservers (running
r151010 although not current on updates) is hanging/locking up
mysteriously (other identical servers run fine). No messages are logged
either to the console or to syslog and the machine becomes almost totally
unresponsive; at
Here's a question: does anyone know of a good way to get detailed NFS
lock information on an OmniOS NFS server, especially information like
which client has a lock on a particular file?
'mdb -k' can be used to extract basic lock information, like the full
paths of all files that have active
| On 2014-04-08 15:44, Saso Kiselkov wrote:
| Anything below OpenSSL 1.0.0 (inclusive) isn't vulnerable to this. (Most
| legacy systems, including OI, still run on the OpenSSL 0.9.8
| release train)
|
| Thanks, I've read that statement ;)
|
| I just wanted to make sure that if we have an
| I am debating the idea of just swapping all my hard drives in my
| current 8x2TB RaidZ2 (all be it slowly) and let the environment
| resilver each drive than expand versus creating a new RaidZ2 on a
| different box and cloning the data over.
|
| Obviously i know of the Pros/Cons/Risks associated
| I know the drive itself does 512b emulation but i would rather run 4K
| if theres a performance increase!
What matters for OmniOS is what the drive reports as. If it reports
honestly that it has a 4k physical sector size, ZFS will say 'nope!'
even if the drive will accept 512b reads and
| In short: when reproducing the bug, try something like vmstat 1 in a
| separate SSH shell, to see if your available memory plummets when you
| disconnect the devices and/or the sr (scanrate, search for swapping)
| increases substantially.
'vmstat 1' shows no sign of this. sr is flatlined at
| On 2014-03-07 21:49, Chris Siebenmann wrote:
|In at least OmniOS r151008j, the iSCSI initiator and thus any iSCSI
| disks it has established are shut down relatively early during a shutdown
| or reboot. In specific they are terminated before halt et al runs
| '/sbin/bootadm -ea update_all
I wrote:
| As far as I can tell from simply looking at things right now, even
| an orderly shutdown on an OmniOS system will not avoid this.
I should clarify that: 'on a normal, stock setup OmniOS system'.
You can of course add SMF jobs to import and export ZFS pools and then
shim them into the
In at least OmniOS r151008j, the iSCSI initiator and thus any iSCSI
disks it has established are shut down relatively early during a shutdown
or reboot. In specific they are terminated before halt et al runs
'/sbin/bootadm -ea update_all' (in halt.c's do_archives_update()).
Under some
| I think a zpool list can help in your debugging to see if the pools
| in question are in fact imported before zfs mount -a, or if some
| unexpected magic happens and the zfs command does indeed trigger the
| imports.
Sorry for not mentioning this before: a 'zpool list' before the 'zfs
mount
I will ask my question to start with and then explain the background.
As far as I can tell from running truss on the 'zfs mount -a' in
/lib/svc/method/fs-local, this *does not* mount filesystems from pools
other than rpool. However the mounts are absent immediately before it
runs and present
| You mention 'directories' being empty. Does /fs3-test-02 contain empty
| directories before being mounted?
It doesn't. All of /fs3-test-01, /fs3-test-02, /h/281, and /h/999
are empty before 'zfs mount -a' runs (I've verified this with ls's
immediately before the 'zfs mount -a' in
| On Fri, 28 Feb 2014, Chris Siebenmann wrote:
| This question makes me feel silly but I'm lost in a confusing maze of
| documentation for sacadm, pmadm, and so on and I can't find anything
| with web searches. What I would like to do is configure what I believe is
| /dev/term/c ('ttyS3
Back in October there was a thread about OmniOS systems hanging on
reboot (starting at
http://lists.omniti.com/pipermail/omnios-discuss/2013-October/001494.html).
I now have a OmniOS r151008j install that does this reliably
on a SuperMicro X9SRH-7TF motherboard (and the workaround from
When I tested and experimented with OmniOS last year, I used pkgsrc.org
as a general source of additional packages. However it currently seems
to be fairly non-functional on OmniOS (or in general, eg its install
instructions point to things that don't exist right now).
Is there some generally
| Also, the ZFS pool backing the NFS server may be slow to write (all
| NFS I/O is sync, and you may require an SSD log device to speed this
| up - seek dtrace scripts that would help you analyze beforehand if you
| have any sync IO that may be the culprit, or temporarily disable sync
| I/O on the
65 matches
Mail list logo