OpenSolaris derived ZFS NAS/ SAN (OmniOS, OpenIndiana, Solaris and napp-it)

The only problem I am aware with new disks on such a config are new disks without a valid partition. But this problem is solved after a napp-it disks-initialize.
I would format/use one of these disks on another computer/pc/mac and retry or try another disk on this server to decide if the problem is on disk side or on server side.
 
AGRH, NFS is eating my nerves...

I'm trying to mount a NFS share in linux. I set share on Omnios:
vmxnet3s0/v4 static ok 10.10.12.200/24

zfs get sharenfs zpool1/gallery
NAME PROPERTY VALUE SOURCE
zpool1/gallery sharenfs root=10.10.12.20:10.10.12.22,rw=10.10.12.20:10.10.12.22 local


Then I try to mount it on the 10.10.12.22 box:
root@cube:~# ifconfig eth0 | grep 10
inet addr:10.10.12.22 Bcast:10.10.12.255 Mask:255.255.255.0

mount 10.10.12.200:/zpool1/gallery /mnt/gallery
mount.nfs: access denied by server while mounting 10.10.12.200:/zpool1/gallery


What am I doing wrong?

Matej
 
OK, if I change sharenfs to root=10.10.12.20:10.10.12.22,
then I can mount it and all new files have UID and GID of 4294967294.

If I change sharenfs to rw=10.10.12.20:10.10.12.22,
then I get the access denied error...

Weird!
 
Hi guys.

After having installed napp-it on a new OmniOS installation, I have a strange problem that I do not fully understand. I hope you guys can point me in the right direction.

I have installed OmniOS and napp-it according to Gea's guides. I also performed the "passwd root" command in order for things to work out with SMB sharing. As I finished my installation, I noticed multiple options in Grub, the first of which was named like my original OmniOS install and the last of which was named "napp-it(version)". I assume that these are so-called snapshots created in the process of installing napp-it, but please correct me if I am wrong.

The problem that I am having is that if I "boot up in"the last one on the list, the one named "napp-it(version)", napp-it works and its web interface is accessible - however, I cannot login to console as root. I have tried enabling root login through the SSH option in the web interface, but that doesn't help. I have also added new local users, but they fail to login as well - with the message "no home directory, logging in with home=/". Also, I cannot login as root when connecting through SMB.

If I reboot and "boot up in" the first one on the list, named like the original OmniOS image, my root login works. Also, the napp-it web interface works, but the configuration is different and only the default pool is to be seen.

Can someone please help a newbie here and tell me, or give me a hint, as to what is going on? If there is a fundamental problem with the installation, I need to know now before I move 10TB of data :)
 
only some hints

- On Solarish systems like OmniOS, you have ZFS on your bootdisk with bootable snapshots,
(they are writable clones of a ZFS snapshot)

- you can create such a bootable snapshot (BE, boot environment) at any time.
This creates a boot entry as well

- if you boot into such a BE, you are not in the state, when you have created the snap
but in the state on shutdown this BE

- you can activate such a BE what means that you boot into per default

After setup, napp-it creates such a snap (BE), activates this BE and you should reboot.
If you boot into this BE, you get always the last state prior reboot

If you boot into the former state, you have the state after setup but prior last actions (like passwd root)

- if you create your own BE to have the option to preserve the current state, you must reboot afterwards
 
Last edited:
Thanks for the reply Gea.

If I understand you correctly, the system must boot in the napp-it BE. But when booted in the napp-it BE, shouldn't I be able to login as root in the console provided that I have allowed root login?

If I did the "passwd root" in the wrong BE (that is, not in the latest napp-it one, but in the), my root password in the napp-it BE should still be blank as it was after a clean OmniOS install. I cannot login as root, neither with blank password nor with the password I set with "passwd root".
 
Last edited:
Thanks for the reply Gea.

If I understand you correctly, the system must boot in the napp-it BE. But when booted in the napp-it BE, shouldn't I be able to login as root in the console provided that I have allowed root login?

If I did the "passwd root" in the wrong BE (that is, not in the latest napp-it one, but in the), my root password in the napp-it BE should still be blank as it was after a clean OmniOS install. I cannot login as root, neither with blank password nor with the password I set with "passwd root".

During napp-it setup via wget
- installer creates a BE (state prior installation) without activating this state
- installer installs tools and configures the ZFS server
- installer creates a BE (state after setup) and activates this state

You must reboot then. It boots automatically into the state after setup
Now you can passwd root and this passwort is stored in this BE.

Everytime when you boot into this BE, you get the last state of this BE
(not the state when creating the BE)

If you cannot login via blank password nor with the password you have entered,
check for uppercase/lowercase problems or keyboard language settings.

If you have no success, try the BE prior installation and reinstall napp-it,
reboot and set a new password after reboot
 
Last edited:
Thanks for the reply _Gea.

I don't think that I am having uppercase/lowercase or keyboard language problems, so perhaps a reinstall is the easiest way to go.

The BE that contains the state prior installation is the "pre-napp-it" one, right?
 
Anyone has an idea about my problem?

Also, when I can mount it, some folder have 40775 permissions or: rwxr-xr-x+. What does the + in the end stands for?

MAte
 
Anyone has an idea about my problem?

Also, when I can mount it, some folder have 40775 permissions or: rwxr-xr-x+. What does the + in the end stands for?

MAte

The + indicates that there are additional ACLs
use napp-it acl extension or /usr/bin/chmod to display ACL
 
AGRH, NFS is eating my nerves...

I'm trying to mount a NFS share in linux. I set share on Omnios:
vmxnet3s0/v4 static ok 10.10.12.200/24

zfs get sharenfs zpool1/gallery
NAME PROPERTY VALUE SOURCE
zpool1/gallery sharenfs root=10.10.12.20:10.10.12.22,rw=10.10.12.20:10.10.12.22 local


Then I try to mount it on the 10.10.12.22 box:
root@cube:~# ifconfig eth0 | grep 10
inet addr:10.10.12.22 Bcast:10.10.12.255 Mask:255.255.255.0

mount 10.10.12.200:/zpool1/gallery /mnt/gallery
mount.nfs: access denied by server while mounting 10.10.12.200:/zpool1/gallery

Some more info:
OK, if I change sharenfs to root=10.10.12.20:10.10.12.22,
then I can mount it and all new files have UID and GID of 4294967294.

If I change sharenfs to rw=10.10.12.20:10.10.12.22,
then I get the access denied error...

Gea, have you ever had such problems?

Matej
 
Can I team network interfaces easily from free version of esxi? I have cheap HP switches, but they do support LACP. All the vmware documentation I can find talks about a networking option I don't have.

edit: never mind. I can do IP hash load balance and non-LACP trunk in my switch. I'm afraid to do mess with it remotely though.
 
Last edited:
Gea, have you ever had such problems?

Matej

No, seems correct, even the /24 in the class A net
that gives you last digit of 1-254

Is your brodacast 10.10.12.255 ?
 
Last edited:
Yes, my broadcast is 10.10.12.255...

Is there a way to enable debuging of NFS in OmniOS? Maby I can get some info from debug. In linux, there is no problem turning up debug, but I dont know what to do in OmniOS...

But get this, if I change my server's IP from .22 to .20, I can mount with rw=10.10.12.20.

Matej
 
Great, will do...

Also, I tried the latest OmniOS the other day and it had bunch of problems. ESX needed 2 minutes to mount the NFS volume:(
And all NFS volumes didn't mount with no_root_squash, even though I set root= for sharenfs....

Matej
 
Great, will do...

Also, I tried the latest OmniOS the other day and it had bunch of problems. ESX needed 2 minutes to mount the NFS volume:(
And all NFS volumes didn't mount with no_root_squash, even though I set root= for sharenfs....

Matej

Create a reverse DNS entry for the vmkernal IP that mounts NFS datastore or put the IP in hosts file.
 
Great, will do...

Also, I tried the latest OmniOS the other day and it had bunch of problems. ESX needed 2 minutes to mount the NFS volume:(
And all NFS volumes didn't mount with no_root_squash, even though I set root= for sharenfs....

Matej

Create a DNS/reverse entry for the vmkernal IP that mounts NFS datastore or put the IP in hosts file.

This is my ISO datastore on OmniOS, no problem at all.
data_download_ISO /data/download/ISO nfs sec=sys,[email protected]/16,[email protected]/16
 
Hum, I tried setting root and rw variables for the vmkernel IP, but it still doesnt work on the new version.

I did also try dns/reverse DNS, but still no luck.

I will probably try with fresh reinstall of latest ESXi and OmniOS...

By the way, which version of ESXi are you using, 5.0 or 5.1? I heard 5.1 has some problems on certain systems...

Matej
 
Just a quick follow-up: I solved my root password issues by performing a re-install. I don't know what I managed to do wrong at during my first install.
 
Hum, I tried setting root and rw variables for the vmkernel IP, but it still doesnt work on the new version.

I did also try dns/reverse DNS, but still no luck.

I will probably try with fresh reinstall of latest ESXi and OmniOS...

By the way, which version of ESXi are you using, 5.0 or 5.1? I heard 5.1 has some problems on certain systems...

Matej

we use 5.1 but had no problems with 5.0 and NFS as well
 
What is the consequences of changing the 'napp-it' UID/GID from what it was installed with (in my case 1001) to a UID > 500. I have lots of users already created from an old zfs dataset and they have UIDs 1000+. I'm running OmniOS v11 r151006 and Napp-It 0.93b from Sept 10 2013. Or is there a way to define the UID for Napp-It at install time? Thanks!
 
What is the consequences of changing the 'napp-it' UID/GID from what it was installed with (in my case 1001) to a UID > 500. I have lots of users already created from an old zfs dataset and they have UIDs 1000+. I'm running OmniOS v11 r151006 and Napp-It 0.93b from Sept 10 2013. Or is there a way to define the UID for Napp-It at install time? Thanks!

no problem.
if you recreate user napp-it, you can set any uid with any pw.
User napp-it should not be able to login (no shell). The account is only used for sudo and file ownership.
 
no problem.
if you recreate user napp-it, you can set any uid with any pw.
User napp-it should not be able to login (no shell). The account is only used for sudo and file ownership.

I did a:

rolemod -u 500 napp-it

Then I had to:

chown -R napp-it:root /var/web_gui

Then I tried to access the web interface. I was getting strange errors related to permissions:

Who are you 1001?

So I did an:

/etc/init.d/napp-it restart

And that seemed to fix the issue. Thanks, Gea_.
 
I'm now trying to get my CIFS working correctly. It's not going well. I read on this thread that the "tank" pool is where authentication polls first. I was trying:

Code:
/usr/bin/chmod -R A- /tank/

And I got thousands of this error:

Code:
chmod: ERROR: Read-only file system

What am I doing wrong? Should I just let the whole thing complete before moving on to:

Code:
/usr/bin/chmod -R A=everyone@:list_directory/read_data/add_file/write_data/add_subdirectory/append_data/read_xattr/write_xattr/execute/delete_child/read_attributes/write_attributes/delete/read_acl/write_acl/write_owner/synchronize:file_inherit/dir_inherit:allow /tank/

I've got NFSv3 working on my OmniNAS now, and it's fantastic. But I absolutely need to get CIFS working too. What am I doing wrong?

EDIT: After a reboot, I have working CIFS. I guess I forgot to reboot after re-applying the root password. However, I'm still encountering the chmod error above. Is chmod choking on my (numerous) snapshots??
 
Last edited:
if you get a readonly error, you have either readonly filesystems below tank or
you have set snapshots to visible (zfs property)
 
I do have snapshots set to visible. Glad its not a major misconfiguration. Thanks, _Gea!

Can anyone explain the CLI difference between CIFS share level ACLs and folder level ACLs? I can see the different links in napp-it, but I want to set these ACLs via the CLI. I'm assuming that the share ACLs are like the Windows share ACL (Everyone = Full). However, I don't understand the difference on the CLI. Are they simply the difference between trivial and non-trivial ACLs?

Code:
# /bin/ls -lVd /tank/public
drwxrwxrwx  11 joltman  public        11 Sep 15 00:23 /tank/public
                 owner@:rwxp--aARWcCos:-------:allow
                 group@:rwxp--a-R-c--s:-------:allow
              everyone@:rwxp--a-R-c--s:-------:allow

# /bin/ls -lVd /tank/joltman
drwx------+ 14 joltman  joltman       19 Sep 14 20:29 /tank/joltman
           user:joltman:rwxpdDaARWcCos:fd-----:allow
                 owner@:rwxpdDaARWcCos:fd-----:allow

# /bin/ls -lVd /tank/wife
drwxrwxr-x+  6 wife  public         8 Feb 24  2013 /tank/wife
           user:wife:rwxpdDaARWcCos:fd-----:allow
                 owner@:--------------:-------:deny
                 owner@:rwxp---A-W-Co-:-------:allow
                 group@:--------------:-------:deny
                 group@:rwxp----------:-------:allow
              everyone@:-w-p---A-W-Co-:-------:deny
              everyone@:r-x---a-R-c--s:-------:allow

Are the ACLs with the @ the "share level" ACLs? I want to set the perms so that wife and I can both connect to public. I should be able to connect to my share, and she to hers. But I want to enter user/pass for all CIFS shares. Then I can map things via My Computer on Windows 7 Pro/Ult. The catch is that public/pictures/videos/joltman/etc are also shared via NFS. Those NFS shares are working perfectly right now. MythTV can record, my HTPC has access to the video/music/pictures shares. I don't want to lose that with my CIFS config.

I've been working on this for a several days now, and I must be missing something obvious. Thanks for putting up with my constant questions!
 
Hi, just a question please :), i've recently reinstalled Openindiana as i had somete issues and can't recall how i did one thing.

So i have set my pool with 755 permissions pretty much in the acl shares and i want a user for xbmc and guests who come to my house to be able to connect to the shares in smb without doing any harm to my stuff, so i have acl shares as 755 and do i just useradd guests and then passwd guests ? although would rather not have that user be able to login into the box but i guess there's no way around it? any way i can set the password for that user just for smb and not be able to login into terminal etc?

Thanks in advance :), is this the right way to do it or would user mapping be easier?
 
I do have snapshots set to visible. Glad its not a major misconfiguration. Thanks, _Gea!

Can anyone explain the CLI difference between CIFS share level ACLs and folder level ACLs? I can see the different links in napp-it, but I want to set these ACLs via the CLI. I'm assuming that the share ACLs are like the Windows share ACL (Everyone = Full). However, I don't understand the difference on the CLI. Are they simply the difference between trivial and non-trivial ACLs?
!

Share level ACL (work like Windows share ACL) are ACLs on the share control-file
/tank/data/.zfs/shares/data (pool=tank, ZFS=data)

Trivial ACL line owner@, group@ or everyone@ are ACL that work
like traditional Unix permissions but with ACL inheritance
 
Hi, just a question please :), i've recently reinstalled Openindiana as i had somete issues and can't recall how i did one thing.

So i have set my pool with 755 permissions pretty much in the acl shares and i want a user for xbmc and guests who come to my house to be able to connect to the shares in smb without doing any harm to my stuff, so i have acl shares as 755 and do i just useradd guests and then passwd guests ? although would rather not have that user be able to login into the box but i guess there's no way around it? any way i can set the password for that user just for smb and not be able to login into terminal etc?

Thanks in advance :), is this the right way to do it or would user mapping be easier?

You can create a user like
useradd -c 'xbmc user' -s /bin/false xbmc

(User cannot login at console because his shell is /bin/false - not working)
 
I've got a speed issue now.

In HD Tune running in a VM (win 7) I'm getting a transfer rate of min 0.4, max 299.9 with average of 110. For the first 50% of testing, speed was measured between 2 and 50. When I did a benchmark in napp-it I get
statfile1 15941ops 266ops/s 0.0mb/s 0.0ms/op 9us/op-cpu [0ms - 7ms]
deletefile1 15929ops 265ops/s 0.0mb/s 7.9ms/op 40us/op-cpu [0ms - 1978ms]
closefile3 15941ops 266ops/s 0.0mb/s 0.0ms/op 3us/op-cpu [0ms - 14ms]
readfile1 15941ops 266ops/s 34.2mb/s 36.9ms/op 37us/op-cpu [0ms - 2321ms]
openfile2 15941ops 266ops/s 0.0mb/s 0.0ms/op 12us/op-cpu [0ms - 65ms]
closefile2 15941ops 266ops/s 0.0mb/s 0.0ms/op 3us/op-cpu [0ms - 7ms]
appendfilerand1 15941ops 266ops/s 2.1mb/s 80.6ms/op 48us/op-cpu [0ms - 2385ms]
openfile1 15967ops 266ops/s 0.0mb/s 0.0ms/op 13us/op-cpu [0ms - 22ms]
closefile1 15967ops 266ops/s 0.0mb/s 0.0ms/op 3us/op-cpu [0ms - 42ms]
wrtfile1 15967ops 266ops/s 33.5mb/s 54.4ms/op 60us/op-cpu [0ms - 3116ms]
createfile1 15991ops 266ops/s 0.0mb/s 5.4ms/op 47us/op-cpu [0ms - 1981ms]
1510: 72.052:

IO Summary:
175467 ops, 2924.193 ops/s, (266/532 r/w), 69.9mb/s, 317us cpu/op, 61.7ms latency
1510: 72.052: S

What can I do to speed things up? Currently just have 5 drives in a raid z1.
 
Solaris 11 server crashed this evening, now on boot the following error appears:

http://i.imgur.com/cqZG6KC.png

Anyone seen this before? I have no idea what it means...

Edit: I can boot into maintenance mode, and can force this crash just by typing zpool status :(
 
rpool is on a 2TB drive and shows 2% capacity


edit: running zdb -emm tank results in the following:

http://i.imgur.com/19VLuf5.png

running the same command on rpool completes successfully

edit2: got it up and running with the pool mounted in readonly... think it may have suffered this bug: http://blog.simplex-one.com/?p=199

any ideas for recovery, or is my only option to back it all up and rebuild?
 
Last edited:
rpool is on a 2TB drive and shows 2% capacity


edit: running zdb -emm tank results in the following:

http://i.imgur.com/19VLuf5.png

running the same command on rpool completes successfully

edit2: got it up and running with the pool mounted in readonly... think it may have suffered this bug: http://blog.simplex-one.com/?p=199

any ideas for recovery, or is my only option to back it all up and rebuild?

If you have payed for Oracle Solaris, ask support if there is a solution available

otherwise, best option seems
- mount pool readonly
- backup
- recreate the pool without L2Arc device
- restore
 
Hey, i was interesting in downloading one of the to go images but they seem to be point to 192.168.1.x could you update please _Gea ? :) thank you very much.

In here : http://napp-it.org/manuals/to-go.html

There is a problem with the download server.
I copy the sample files over to a new machine. This can last same time.

ps
The USB image tools are available for download.
You can installl OmniOS with all needed tools and settings to an USB stick and
use the USB image tools to backup/retsore/clone your own customized napp-it To Go sticks.
 
Back
Top