Wednesday, December 21, 2011

EON ZFS Storage 1.0Beta based on oi 151a release!

I promised everyone patiently awaiting the new release, a pre Dec 25th gift and HERE IT IS!!! Lots of hard work went into giving this release life. It has undergone many hours testing so I could deliver another "rock solid" release. This build, required 2 virtual machines, 3 physical test machines and 3 compilers. Please enjoy and feel free to post feedback or any anomalies encountered. I hope EON continues to store all your precious digital moments!










New to EON, read here first! Upgrading, read here first!
For new setups, before running "setup":
cd /mnt/eon0/bin
./slinky r
./slinky c
setup
If your EON Storage is internet connected. The EON ISO can be retrieved using wget.
After running setup(new setup) or transporter.sh (upgrade/backup). Re-check/edit your POOL=poolname in /mnt/eon0/.exec.
export POOL=YOUR_POOLNAME
This is an option to edit "YOUR_POOLNAME" with a simple command if you wish to avoid "vi" (substitute your pool name where "YOUR_POOLNAME" appears). A backup file will be saved as .exec.bak
cd /mnt/eon0
sed -i.bak 's|=abyss|=YOUR_POOLNAME|g' .exec
An alternate upgrade option, using a spare USB key. Perform a new setup and then a forced zpool import to test. Perform the necessary tests and make sure all is ok before running:
zpool import -a -f
updimg.sh /mnt/eon0/boot/x86.eon
After running the new install for a few days and triple checking everything is ok. Only zpool/zfs upgrade is left, to access the new features of zpool version 28 and zfs version 5. Note that once zpool/zfs upgrade is done, there is NO GOING BACK to the older version.
zpool upgrade poolname
zfs upgrade -r poolname

EON 64-bit x86 CIFS ISO image version 1.0b (NO HTTPD)
EON 64-bit x86 Samba ISO image version 1.0b (NO HTTPD)
EON 32-bit x86 CIFS ISO image version 1.0b (NO HTTPD)
EON 32-bit x86 Samba ISO image version 1.0b (NO HTTPD)
Binary Kit 1.0:
  • binary kit 1.0 (Released 22-February-2012)
  • rail (remote internet package installer)
  • emp (EON multi-purpose program)
Pre-requisite: If you had a previous binary kit installed
cd /your_zpool
mv local old-local
mkdir local
If this is the first time using the binary kit
cd /your_zpool
mkdir local
Perform the following on the EON command line to install the latest binary kit, remote installer and more. Most of these steps are one time only to add the bits that were not ready at release time. They will be self updating from here on. You can simply copy and paste 1 line at a time in an ssh session. Run "rail" to see more usage options.
cd /mnt/eon0/bin
./slinky wget
wget -q -U mozilla -O slinky http://bit.ly/zNY7rI
wget -q -U mozilla -O md5sum http://bit.ly/xwQZU3
wget -q -U mozilla -O /usr/bin/rail http://bit.ly/yqck53
chmod 755 md5sum slinky /bin/rail
./slinky md5
rail install emp
rail install binkit
rail install perl
rail
Web Server: Installing a web server (apache, lighttpd or nginx). Browse to your IP after starting a web server to see links to various apps. Run "emp" to see usage options. More detailed how to here.
rail install apache
certgen.sh (to generate unique self-signed certificate) 
emp apache start
Transmission: Edit /usr/local/transmission/settings.json to allow your client IP before starting Transmission Bit Torrent client
rpc-whitelist": "127.0.0.1,IP_address_of_client",
rpc-whitelist": "127.0.0.1,10.1.1.5",
emp transmission start
Python is a requirement for SABnzbd, Sickbeard and CouchPotato.
rail install python
SABnzbd: Installing and starting SABnzbd. More detailed how to here.
cd /tmp
wget -O sabnzbd.tar.gz http://sourceforge.net/projects/sabnzbdplus/files/sabnzbdplus/sabnzbd-0.6.15/SABnzbd-0.6.15-src.tar.gz/download
gzip -dc sabnzbd.tar.gz | tar -xf -
mv SABnzbd-0.6.15 /usr/local/sabnzb
emp sabnzb start
Sickbeard: Installing and starting Sickbeard. More detailed how to here.
cd /tmp
wget -O sickbeard.tar.gz --no-check-certificate https://github.com/midgetspy/Sick-Beard/tarball/master
gzip -dc sickbeard.tar.gz | gtar -xf -
mv midgetspy-Sick-Beard-911f5d1 /usr/local/sickbeard
emp sickbeard start
CouchPotato: Installing and starting CouchPotato. More detailed how to here.
cd /tmp
wget -O couchpotato.zip --no-check-certificate https://github.com/RuudBurger/CouchPotato/zipball/master
unzip -q -aa couchpotato.zip
mv RuudBurger-CouchPotato-6970c10 /usr/local/couchpotato
emp couchpotato start 
New/Changes/Fixes:
  • New ZPOOL (version 28) and ZFS (version 5)
  • Perl 5.10.0 (full version, user customizable) moved to the binary kit, requires a new setup step (slinky)
  • New Binary Kit is semi-mandatory (needed for S.M.A.R.T and more)
  • New/Updated administration and statistics binaries.
  • Simple appliance package downloader and installer (more info to come)
  • kernel line addition "-B disable-pcieb=true" under VMware. ESXi, Fusion, no longer required.
 S T O R E   I T   Y O U R   W A Y !

75 comments:

durian said...

Thank you! can't wait to try it out.

Somerandom said...

Hi Andre,

Nice work!
Will 32 bit builds be coming online later as well?

Cheers,
John

Max said...

It really is a special time of year! A Newegg gift is en route! :-)

Thanks a bunch!

danny said...

Hi Andre,

Thanks for eon! Since two years ,I
am working with eon on my nas and
i frequently check your blog for new releases with new features. Today I installed the new io-beta and i found some unknown messages ,when i create a zpool-volume and zfs.

SMF Initialization problems..svc:/network/nfs/server
SMF Initialization problems..svc:/network/nfs/nlockmgr
SMF Initialization problems..svc:/network/nfs/client
NFS plugin problem with SMF repository: unknown -1
so far.
But the creation seems to work.
Is this harmful?

Thanks alot and Merry Xmas
Danny

Andre Lue said...

Hi Danny,

Those messages are harmless. They usually occur at zpool import when you import a disk previously used with zfs that had a service which is currently not running.

In the case below it looks like NFS was enabled but it is not enabled on the current system. I will look into this further.

Andre Lue said...

Hi Max,

Thanks, much appreciated! Many of the items on my newegg list/links are out dated. I will have to get them updated/cleaned up.

Hi John,

Yes, 32 bit builds will come online later. I will try to make it before the Dec 25th but not sure I'll have them up before.

S Javeed said...

André,

You are the man! Thank you for providing upgrade instructions that won't force me to upgrade my zpool just yet :-)

Shahbaz

Jim Doyle said...

Looks good. Tried it out abit. Dont forget to add the SUN C++ runtime shared libs to the default build that Java 6 and Java 7 will fire up:

# ls /usr/lib/libC* /usr/lib/amd64/libC*
/usr/lib/amd64/libCrun.so.1 /usr/lib/libCrun.so.1
/usr/lib/amd64/libCstd.so.1 /usr/lib/libCstd.so.1
/usr/lib/libC.so.5

Novell Guy said...

thanks for all your hard work at the holidays...

Andre Lue said...

Hi Jim,

The binary kit will provide all those bits and more :)

Roenbaeck said...

The zpool import -a -f fails for me. First I get a lot of "SMF Initialization problems" which seem to be NFS related, followed by "NFS plugin problem with SMF repository: unknown -1". Those are probably related. However, after that error it says "cannot import 'terapool': invalid vdev configuration", where 'terapool' is the name of my pool (2TB 2 disk mirror).

Roenbaeck said...

Sorry, it's a 3TB pool with one 2TB 2 disk mirror and one 1TB 2 disk mirror.

Andre Lue said...

Hi Roenbaeck,

The SMF, NFS errors are harmless (see Danny's post above). They usually mean you had NFS services on the pool being imported and the NFS services are not started/running on the current machine doing the import. After enabling NFS those messages will go away.

Regarding the zpool import failing. Sounds like one pool imported ok and one failed. Kindly email me:
zdb -v
zpool status

Have you tried importing that pool by pool id? Share any important specifics that would help regarding how the pool was built. Are the 2TB disks same model, brand hard drive etc? My early guess is something in the mirrored pair's format or label differs and one drive's partition is too small.

Roenbaeck said...

Could not find your email, so I am posting it on my blog. Fortunately I documented my entire experience with EON Storage so far here:

http://doitopen.blogspot.com

You can see how I set up everything there, including the output from the commands you requested.

Andre Lue said...

Hi Roenbaeck,

I see! Very interesting what happened here. The set of commands used to build your pool should have been done with all same size disks, yielding a raid-10 equivalent (ie since it started with a 2x2TB mirror, the "zpool add mirror ..." should have been another pair of 2TB disks not 1TB). The commands should not have allowed a 2x1TB mirror be added to a 2x2TB mirror. Did you use -f switch to get it added?

Also, the Sept post shows (c1t0d0p0 c1t1d0p0)but the current listing shows c2tXdXpX. Using a different machine or controller?

My recommendation is to BACKUP your data and attempt to detach the 2x1TB pair. If that succeeds, the 2x2TB mirror should import fine and can add another 2x2TB later. You can always build a separate pool with the 2x1TB.

Roenbaeck said...

I did not have to use the -f switch, and I thought the whole idea of a pool was that you can add differently sized raid-sets to it?

At some point I attached a USB disk to the system, and I believe that must have been when the controller numbering changed from c1 to c2.

If I can detach the second (smaller) mirror without any warnings, does that mean there was no data loss, and no restore will be needed from the backup?

Andre Lue said...

Hi Roenbaeck,

I was able to reproduce building a similar pool, which seems like a bug. I am not sure of a clean way out of this that doesn't involve destroying and rebuilding the pool.

Because, you would have to attempt to detach both 1TB disks, I'd say there's a high risk of data loss. I'll have to think about this some more.

Roenbaeck said...

Thanks for looking into this. I'm currently backing up everything (2.5TB) and it will still take another couple of days.

Andre Lue said...

Hi Roenbaeck,

On further research, the pool and the steps used to build it are ok. The problem seems to be related around these being 4k sector drives with logical 512 presentation. The zpool version (snv_130) used to build has no 4k sector checks, while the latest release (oi 151) has some checking but not a real 4k sector check. Unfortunately, until the zfs 4k sector check is fixed, the import will halt with the "invalid vdev configuration" message for appended pools using different size 4k sector drives. If these were the same size I think this would not have the same issue. I'll keep you posted on my findings.

Denis J. Cirulis said...

Hi Andre,
good to see the news on EON!
Just wanted to know are there any changes on adding drivers and opnindiana components to EON image ?

Noticed that I need mptsas driver for my enclosure and srpt package as well, to connect to my nas via infiniband. What is the short way to do it ?

Jon Strabala said...

Hi Andre,

Very interested in EON especially now that you have a 151a setup. But I have two questions.

#1 Can I compile EON form source (or add it to my oi_151a base), my hardware always need a special BE to eliminated motherboard/chipset/cpu specific bugs, I always load the oi_151a ISO follow the https://www.illumos.org/projects/illumos-gate/wiki/How_To_Build_illumos for recompiling with a few key patches (or even start with the most recent nightly), boot into the new BE delete my original installed BE.

#2 Is there any support for a HA EON system e.g. syncronizing file systems across two EON servers with a floating IP address ?

Thanks in advance

Jon

Andre Lue said...

Hi Denis,

Infiniband and srpt drivers support are available in a commercial offering of EON. Please feel free to email me for more info.

Hi Jon,

1) Give the current release a boot/try and see if it works. Report back any errors or what doesn't work.

2) HA cluster + EON is still in the early phase. Still working on a clean, error free compile. When finalized HA + EON will most likely be a commercial offering.

us said...

@andre
great work

using 64-cifs

I did a new setup, after importing zpools as written I had to export them all once so "zpool import -a" in .exec works w/o -f switch.

also getting a lot of "SMF init... probs..."

and I also getting every boot:

WARNING: No randomness provider enabled for /dev/random. Use cryptoadm(1M) to enable a provider.

an old known error, see here
https://www.illumos.org/issues/995 or
http://defect.opensolaris.org/bz/show_bug.cgi?id=38


another problem is "init 6" fails to do the rebooting at a 100% failrate (w/ b134 only 20-30% failrate), "init 5" also fails to power off the machine, but with this the successrate wasn't good either with b134.

Andre Lue said...

Hi us,

The SMF, NFS errors are harmless (see Danny's post above). They usually mean you had NFS services on the pool being imported and the NFS services are not started/running on the current machine doing the import. After enabling NFS those messages will go away.

The "WARNING: No randomness provider enabled for /dev/random. Use cryptoadm(1M) to enable a provider." message is harmless. It's a boot race condition between ZFS and swrand. It could probably be fixed by a SMF to wait until swrand provider is online but it would extend boot wait time.

Regarding init 6 failing, I've never really seen this, not at the rate you mentioned. Which model/brand motherboard is this?

I've seen init 5 failing to power off the machine which is usually due to a mismatch between acpi bios and illumos/oi OS implementation. You can try booting the acpi options at EON boot screen to troubleshoot further.

Aaron aka Frank said...

Hi Andre

Not to be pushy or anything (and sorry if it seems I am) but any progress on the 1.0beta binary kit? I'm eager to fire up Sickbeard but the version of Python I have (which I grabbed from the now defunct IPS repo) is too old to run it. Might fire up snv_130 in a VM and build it from there (lucky I grabbed an iso before Oracle destroyed it!).

Thanks, and once I have the means I'll be sure to throw an Amazon gift voucher your way :)

Andre Lue said...

Hi Thirdwheel,

I was trying resolve some deprecated calls for mediatomb and mt-daapd against the latest ffmpeg. I will fall back to ffmpeg-r26402 and try to resolve the mediatomb and forked-mtdaap issues with the newer ffmpeg later.

Roenbaeck said...

Any news on the 4k sector check? Once it is properly in place, what will happen? Since my disks fail that check already now, won't they always?

I am wondering if I should backup, remake the file systems in 151a, and restore.

Jeff said...

the NFS manifest file is not in the /var/svc/manifest/network folder. Trying to enable nfs will always fail.
Is NFS going to be part of binary kit?

Andre Lue said...

Hi Roenbaeck,

It is indefinite when that code will be fixed. I would proceed with the backup and re-creation.

Hi Jeff,

NFS is still there and works, but since Illumos
/var/svc/manifest
has been changed to
/lib/svc/manifest
Change var to lib and all should work as before.

us said...

hi Andre,

I have an MSI workstationboard with 945GT intel chipset. w/ b134 I played around with the acpi-options but no changes, w/ b151 I didn't tried.

LittleMonkeyMojo said...

Been using the 64-bit CIFS build, booting off USB. Appears that /usr/perl5 is symlinked to /usr/local/perl5, and /usr/local/ is symlinked to /abyss/local.

With this, such commands as /usr/bin/kstat fail as the perl binary can't be found.

Andre Lue said...

Hi,

This is due to the change to release a full user modifiable perl in EON 1.0.

Perl will be shortly available as part of the binary kit.

Finally, to connect all the pieces you need to edit your ZPOOL name in /mnt/eon0/.exec
export POOL=your_pool_name

LittleMonkeyMojo said...

System won't re-establish connection with zpool upon reboot. If I manually run the command: zpool import -a
I get the following error:
"cannot import 'hyperion': pool may be in use from other system, it was last accessed by hyperion.si.umich.edu (hostid: 0x7be05) on Mon Feb 20 14:38:48 2012"

Using "zpool import -a -f" works, but complains on one zfs filesystem which has a different mountpoint than name (Name:hyperion/l/local Mountpoint:/usr/local)

Andre Lue said...

Hi LittleMonkeyMojo,

It appears you have the zpool host on the booted system and the last host that mounted the system out of sync.

This is probably due to a missed step during setup. After you import -a -f with a configured system (hyperion) it should be all corrected. If it does not correct it then you most likely missed a step during setup.

The mount probably complains because /usr/local is already reserved for the binary kit.

Alain F. said...

Hi Andre,

I recently upgraded to the 151a release and while I've encountered a few of the issues listed so far, I've managed to resolve them.

I do have one lingering issue in that certain changes made to the system are not preserved on reboot, despite meticulously running updimg.sh

An example is when I create a /home directory for a user, upon reboot the directory no longer exists.

Any suggestions on what's going on?

Andre Lue said...

Hi Alain,

The design behind this, is all other users (non-system, login allowed) home directories would be created in your zpool. The user's home directory can be specified using -d argument. The reasoning is, a user allowed to login will most likely add/create files and require storage.

eg: useradd -d /zpool/username

Chris said...

Hi! I'm trying to get glusterfs running on top of eon. There is a script in the Download section, but how do I use that? Thanks alot for all the work! Chris

Unknown said...

Hallo Andre,

I am trying hard to get the 151 version running, but I face so many issues.

first, the system complains on install about not being able to connect to dyndns, what screws up the dns resolve mechanism.

this leads to errors on your scripts, especially the slinky stuff. I have done some research and found, the wget command overwrites the slinky with a zero length file, because it cannot resolve http://bit.ly/...
I have done the steps by hand giving the ip address instead of hostname.

however I cannot extract the binary package without errors.

the libgcrypt.so.11.7.0 seems to have issues being extracted out of the kit.

I have a 8G fast usb stick, an I would like to have all binaries on that device. The reason is, I do import or export my pools, I do not want to lose my binaries while doing so. I do not care about having ten versions of eon stored on the stick, but would like to have all files in access unpacked, so I can add or remove features easily.

Is there a way to have all the binaries and settings on the flash (if necessary I can buy a bigger stick)? Otherwise zpool maintenance becomes very troublesome...

Unknown said...

...

by the way, I can follow the concept of having the users home directories on a user zpool instead of the system stick, but what is the reason to dump the root's home directory?

This is for special purpose to allow the root to store necessary stuff for administration.

Is it possible to have at least one mount point on the boot device allways RW accessible, no matter what version is booted, to allow to use the remainig space for storing extensions or user scripts?

p.s. the issue with dyndns dissapears after setup / reboot. it seems to be only on the iso image.

Thank you very much

Andre Lue said...

Hi Chris,

Glusterfs was part of the previous binary kit. It was not a part of the current binary kit because there was no feedback, and minimal community interest. I may put it back as it seems RedHat have started updating the project.

Hi xbmcg,

The dyndns warning messages are harmless. Before attempting to install the binary kit, EON should already be booted, setup and running from USB, with working internet access.

Repeating the update of slinky will resolve the zero file once your internet access is fully configured.

I am unaware of a libgcrypt error. Can you detail what/how this error could be reproduced?

It is possible to have the binary kit on the flash but beware there are many programs included that log continuously (apache, transmission, etc) and could cause early failure of the flash based on frequency of writes. I do not recommend doing this.

Can you explain how zpool maintenance becomes troublesome if the binary kit is not on the USB key?

USB install: /mnt/eon0 is automounted and always writeable? Not sure I understand your root RW question? Please clarify.

apocalipsis said...

Error updating configuration:

/usr/bin/updimg.sh /mnt/eon0/boot/x86.eon
EON Storage updater
Updating files in /mnt/eon0/.backup to x86.eon
backup in /mnt/eon0/boot/x86.eon.1
/mnt/eon0/.backup: OK
gzcat /mnt/eon0/boot/x86.eon > /tmp/x86.1330
lofiadm -a /tmp/x86.1330 /dev/lofi/1
mounting /dev/lofi/1 /mnt/upd
OK: mount /dev/lofi/1 /mnt/upd
updating /etc/svc/repository.db -> /mnt/upd/etc/svc/repository.db
umounting /mnt/upd
lofiadm -d /dev/lofi/1
mv -f /mnt/eon0/boot/x86.eon /mnt/eon0/boot/x86.eon.oem
mv: cannot rename /mnt/eon0/boot/x86.eon to /mnt/eon0/boot/x86.eon.oem: Read-only file system
mv -f /mnt/eon0/boot/x86.eon /mnt/eon0/boot/x86.eon.0
mv: cannot rename /mnt/eon0/boot/x86.eon to /mnt/eon0/boot/x86.eon.0: Read-only file system
gzip -f -9 -c /tmp/x86.1330 > /mnt/eon0/boot/x86.eon
/usr/bin/updimg.sh[74]: /mnt/eon0/boot/x86.eon: cannot create [Read-only file system]

Andre Lue said...

Hi Apocalipsis,

It appears you are performing the updimg step with a "Read-only file system"

Kindly re-check your steps.

apocalipsis said...

Yes... that was the problem

Chris said...

Hi Andre,

thank you for your response. I think you should really evaluate the newer versions of gluster as they work very well (on linux). I tried to compile from source on openindiana and it builds fine but fails with some library dependency errors when run. If I get it to work, i'll post it here. I think that extending the capabilities of eon by HA-functionality (not just zfs send) would really boost the level of recognition!


Hi xmbcg,

my steps to get DNS working was to add the server to /etc/resolv.conf and copy over the /etc/nsswitch.dns to /etc/nsswitch.conf. That would be
#echo "nameserver 10.0.0.1" >> /etc/resolv.conf
#cp /etc/nsswitch.dns /etc/nsswitch.dns

Unknown said...

Hello,

thank you for your quick reply.
I have got it one step further.
The crypto lib issue was my mistake.
I have tried to map the abyss pool-local to a writeable folder inside /mnt/eon0/bin, hoping this is something, tat survives reboots - but I was wrong. Actually it was probably because of ram-disk overflow, so the library could not be unpacked. I have unpacked the package in a linux box, and tha worked - so nothing wrong with the package so far...

I do agree, that writing could be an issiue for the binary kit. however I just need 2 progs, nano and / or mc to have a file manager / text editor for patching the config files. all other stuff I do not care about.

The main issue with binary packages / swap space on the zpools is, once you export the pool, you hang-up the system or loose the tools (e.g. to get the drives out of the box or to replace them and re-configure the nas. I would prefer a nas, that is independent functional of attached drives. Any way, great work so far. I just need to understand how to add some extensions anch patch settings, that survive a reboot...

Thank you very much!

Andre Lue said...

Hi xbmcg,

EON is fully independent of the binary kit to perform any drive, export operations or preserve configs beyond reboot. The binary kit only adds application space functionality that users have asked for.

The dependencies for mc are plenty vs the returns, hence it not being included. You can share your findings on mc for others as users may find mc useful.

It's pretty simple to have things preserved beyond a reboot. If you do not want the binary kit on the zpool edit the /usr/local symlink in /mnt/eon0/.exec.

I do not follow what is meant by extensions but if you mean your own binaries you will need a compiler, the open source packages and all the dynamic dependencies the binary requires.

Dave said...

Andre,

I was wondering when the next release would be ready?

I was also wondering if it would be possible to start including 10gbe drivers, specifically ixgbe

Thanks!
Dave

Unknown said...

Hello Andre,

Thank you very much. There is a little too few information on the storage locations of the files and the boot process / save settings / backup procedures - I think I have to get more familiar with the system first. Anyway, it is very impressive, how small you squeeze it! Thank you for the hint with the .exec file. I also have found some more hidden important files, that take care on the directories squeezed into the image. Mainly I come from the windows world, but have also some basic linux knowledge - but using the vi is still a pain. BSD / Solaris is a little more different, so there is a learning curve in front of me... Sorry for some of my stupid questions...

jdef said...

Hi Andre, I'm not sure this is the right place for this comment: I'm still using the non-beta version (0.60). I just set up link aggregation via dladm and a hostname.aggr1 file. Rebooted and its working great. I see errors on bootup though and I think it's because I can't get rid of the old hostname.nge0 file - that I no longer need since that NIC is part of the aggr1 that I set up. I even deleted all the contents of the file and I still get the same error, as if solaris is trying to plumb nge0 even though it's part of aggr1.
How to I get updimg.sh to remove the old hostname.nge0 file?

jdef said...

Looks like my last comment appears to be from "Unknown". Any way, I tried adding the hostname.nge0 file to the .remove file -- and it does seem to get removed by the time I'm able to log in and look at /etc. However, the boot errors persist. I wonder if .remove is executed after the network interfaces are init'd?
Here's what the errors look like:

Mar 2 16:21:27 eon ip: [ID 210580 kern.error] nge0: DL_BIND_REQ failed: DL_SYSERR (errno 16)
Mar 2 16:21:27 eon ip: [ID 954723 kern.error] nge0: DL_UNBIND_REQ failed: DL_OUTSTATE
Mar 2 16:21:27 eon pcplusmp: [ID 805372 kern.info] pcplusmp: ide (ata) instance 0 irq 0xe vector 0x45 ioapic 0x2 intin 0xe is bound to cpu 0
Mar 2 16:21:27 eon pcplusmp: [ID 805372 kern.info] pcplusmp: ide (ata) instance 1 irq 0xf vector 0x45 ioapic 0x2 intin 0xf is bound to cpu 1
Mar 2 16:21:29 eon mac: [ID 435574 kern.info] NOTICE: nge0 link up, 1000 Mbps, full duplex
Mar 2 16:21:37 eon ip: [ID 210580 kern.error] nge0: DL_BIND_REQ failed: DL_SYSERR (errno 16)
Mar 2 16:21:37 eon ip: [ID 954723 kern.error] nge0: DL_UNBIND_REQ failed: DL_OUTSTATE

Andre Lue said...

Hi jdef,

To fix this issue. You will need 2 ssh sessions or 2 enable virtual consoles

cd /tmp
cp -p /bin/updimg.sh .

Edit updimg.sh to add a pause
change lines
#backup
update_file
to
#backup
echo -n "Press enter to continue " ; read x
update_files

Then in one session run
/tmp/updimg.sh

When you see the "Press " msg, in the other (ssh or virtual) session
cd /mnt/upd/etc
rm hostname.nge0

Go back to session 1 and press enter to complete updimg.sh

jdef said...

Thanks, Andre. That killed the file. To anyone else who tries this, be sure to 'cd' outside of /mnt/upd before pressing ENTER to continue otherwise the updimg.sh script won't be able to umount the directory properly.
So, Andre, that killed the "top half" of the DL_XXX error messages that I'm seeing. I suspect that it has to do with either the order in which the devices are being init'd, or else perhaps the nge driver:


Mar 5 12:43:10 eon mac: [ID 435574 kern.info] NOTICE: e1000g0 link up, 1000 Mbps, full duplex
Mar 5 12:43:10 eon mac: [ID 469746 kern.info] NOTICE: aggr1 registered
Mar 5 12:43:10 eon mac: [ID 435574 kern.info] NOTICE: aggr1 link up, 1000 Mbps, full duplex
Mar 5 12:43:10 eon pcplusmp: [ID 805372 kern.info] pcplusmp: ide (ata) instance 1 irq 0xf vector 0x46 ioapic 0x2 intin 0xf is bound to cpu 0
Mar 5 12:43:11 eon pcplusmp: [ID 805372 kern.info] pcplusmp: ide (ata) instance 1 irq 0xf vector 0x46 ioapic 0x2 intin 0xf is bound to cpu 1
Mar 5 12:43:13 eon mac: [ID 435574 kern.info] NOTICE: nge0 link up, 1000 Mbps, full duplex
Mar 5 12:43:21 eon ip: [ID 210580 kern.error] nge0: DL_BIND_REQ failed: DL_SYSERR (errno 16)
Mar 5 12:43:21 eon ip: [ID 954723 kern.error] nge0: DL_UNBIND_REQ failed: DL_OUTSTATE

Thoughts?

Dave said...

Hi Andre,

Any word on the 10Gig drivers? I would also be interested in seeing the Myri10GE driver included.

Thanks!
Dave

Tresten said...

Hi Andre,

Do you see this when you run psrinfo?

bash: /usr/sbin/psrinfo: /usr/perl5/bin/perl: bad interpreter: no such file or directory

Nice work by the way ... I used EON ages ago and now I feel the need to come back!

Regards,

Antony Brooke-Wood

Andre Lue said...

Hi Morph,

I'd have to double check the myri10GE but ixgbe and other 10gigE drivers are available in a commercial version.

Hi Tresten,

Thanks. That message appears because psrinfo is a perl script and perl is now an external add on. The trade off is that Perl is now a full version and user modules can be installed as with a normal install. To install Perl you can use "rail". Please see the binary kit section of the post.

Dave said...

Could you elaborate on this "Commercial" version, I haven't seen or heard of this before, nor have I been able to find a feature comparison.

I also wouldn't be using it in a commercial environment, This would all be at home.

Thanks!
Dave

Andre Lue said...

Hi Morph,

A commercial version has always been available for commercial or any form of business related use.

The difference lies in the drivers included. For example 10gigE drivers are mostly found in a enterprise/production environment. So, those drivers are only included in the commercial version.

Hope that helps.

Dave said...

what would the price be for the commercial version for personal use?

Andre Lue said...

Hi Morph,

Regarding license info, please send an email to eonstore AT gmail DOT com.

Lane said...

Any chance this can run on Citrix XenServer or open source Xen 4?

Andre Lue said...

Hi Lane,

I don't see why it wouldn't work. EON's been tested to work with VMware and VirtualBox.

durian said...

Hi Andre,

Did you have a chance to try Napp-It on the 1.0Beta ?

Thanks

Sami said...

Hi

I have installed this beta, done the setup and updimg.sh and rebooted.
I then created a zpool and a zfs on it.

However, when I reboot the machine, the zpool isn't remounted.
/usr/sbin/zpool import -a
gives
cannot import 'abyss': pool may be in use from other system
use '-f' to import anyway

br
Sami

Andre Lue said...

Hi Sami,

That usually occurs when a step with slinky was missed during setup and /etc/hostid is most likely blank or does not exist.

Perform the following steps to fix:
rm /etc/hostid
cd /mnt/eon0/bin
./slinky r
./slinky c
/lib/svc/method/svc-hostid
cat /etc/hostid

Run updimg.sh to preserve and finish.

Hi Durian,

Please see this thread for more on napp-it.

jdef said...

I needed mpt_sas driver support, so I pulled drivers from the OI 151a iso. YMMV.

harl2000@gmail.com said...

Nice job ! Andre.....

Is it possible to install VMware Tools on EON once is running as a VM on ESXi?

matej said...

Hi,
I was following the Getting started manual and got to this command:
zpool create abyss /dev/ramdisk/ram1

in fact I was using

zpool create zfs /dev/ramdisk/ram1

because I wanted the pool to be named that way. The thing is that I forgot that I have a hard drive connected with a zfs pool (from FreeNAS) on it which has the same name, i.e. zfs. After rebooting zpool list shows me no pools. Is my data on the hard drive lost or what can I do? Is it really that easy to fuck things up? I would have expected it to warn me about these things on zpool create...

Or have I just missed some info on how to correctly import the pools on hard drives?

Thanks very much,
matej

Tehcook said...

Hi Andre,

Do any of the existing pages describe proper way to create binaries for this 1.0 beta ? I'm interested in a newer Netatalk/AFP to support Lion time machine. But not sure if older pages describing IPS and rolling own binaries still apply.

Andre Lue said...

Hi Tehcook,

The latest version 2.2.1 works with LION. If I'm missing something please let me know. Give it a try
rail install afp

There is also no IPS netatalk package as far as I know, so you would have to compile it yourself. I plan to update 2.2.1 but there are higher priorities in the queue.

Tehcook said...

Thanks. I might have been fooled by the Time Machine OS X page in Wiki saying its at ver 2.0.5. Rail does indeed installs 2.2.1 which should be fine.

Does it make any sense to dump rail and emp into /usr/local/bin ? Otherwise they get wiped out after reboot.

Andre Lue said...

Hi Tehcook,

The 2.0.5 is a previous user guide but still a useful read. To preserve emp/rail beyond reboots add the entries below to /mnt/eon0/.backup and run updimg.sh
/usr/bin/emp
/usr/bin/rail
These are intended to update themself (user invoked) as they evolve. A update package containing all the eon bits will be added soon.

rail/emp was not put in /usr/local/bin because rail installs /usr/local/bin and emp can/will be used to start things at boot (hotplugd, etc) before /usr/local/bin is mounted. rail should be fine there though if you choose.

Tehcook said...

Hi Andre,

A question about updimg.sh script. Which was often failing for me. Not a big surprise as there is not a lot of error checking being done. I think the reason it often fails is running out of space in /tmp. Right now my server has :

Filesystem size used avail capacity Mounted on
swap 64M 0K 64M 0% /tmp


-r--r--r-- 1 root root 76M May 30 00:35 /mnt/eon0/boot/x86.eon
-r--r--r-- 1 root root 76M May 29 23:34 /mnt/eon0/boot/x86.eon.0
-r--r--r-- 1 root root 75M Dec 21 10:59 /mnt/eon0/boot/x86.eon.oem

At this point it most likely will fail if I do updimg.sh.

I was thinking of changing updimg.sh to check if local disk pool is available and if so - use it as TMPIMG. Do you foresee any issues with doing test of the local pool mount and if its there - change TMPIMG=/tmp/x86.$$ to TMPIMG=//x86.$$ ?

Andre Lue said...

Hi Tehcook,

Sounds like something is misconfigured or you may have too little RAM.

For best performance you should have a ZFS swap configured. If ZFS swap is not configured, the update uses RAM. With 2GB or 4GB of RAM you should have more than enough to update the image.

For quicker responses you can post to the EON group

NiTRo said...

Hi Andre, any new build since Xmas ? Thanks a lot for your hard work, EON is awesome!

Lobo71 said...

Hi Andre,

First, great job ...
I'm using EON since a couple of weeks and it it a very good SOHO-HOME NAS.

I would like to use sys-suspend.
Unfortunately, it doesn't work because of display (the error message is cannot suspend display).

I'm using a NVIDIA graphic card, how can I make EON using the NVIDIA driver ?

Is there a way to make sys-suspend work ?

Best regards.

José