Can I create a Freenas 9.3 boot USB on different hardware than I will run it on?

[SmilingInSeattle]

I would like to create a new boot USB for an existing Freenas server, but I don't want to pull down the server and move it to monitors and a keyboard to do so.

While the server is running the latest version of Freenas 9.3, with the current update patches, the existing USB stick is getting old. I recently bought two quality USB 3.0 sticks with the intent of mirroring the existing drive, but found they to be too small by about 200MB to do so.

Therefore, I'd like to use another computer to run the latest Freenas 9.3 iso on and create a new Freenas flash drive on it, using the curses menu to force as much of the networking information that I have, save the latest configuration on another machine on the network, detach my zpool, shut down my Freenas server, pull the old OS stick out, put the new one in, reboot the server, and reattach the existing zpool . Once I reboot the server, my plan would be to upload the saved configuration db, put the 'sister' USB stick I just bought in another USB port and then create a mirror for the sake of having a current "failsafe" backup USB OS disk. And now I see one more thing, if that works: buy a couple more of the same brand/model USB sticks as I just bought so I create future failsafe backups!

 

[Bidule0hm]

Yes, you can do that ;)

 

[David Dyer-Bennet]

I've created my FreeNAS flash boot drives on Windows boxes consistently; definitely does not need to be done on the same hardware, it's just a question of having the right software for that OS to do the right thing with the image (put it on the USB the right way I mean).

 

[SmilingInSeattle]

It's no problem to write the iso to a USB stick. What I wondered is if I then take that iso file, boot another computer other than the NAS from that ISO file, write the book image (which since FREENAS 9.3 is a ZFS imzge) to another USB, will the process probe the computer the boot drive is being created from and set some configurations that are not totally generic? The intent is to take that second USB which will be written to a ZFS pool, put it in the NAS and boot the NAS from it.

 

[David Dyer-Bennet]

Okay, I see your issue. And I don't know the answer, that's gone past my detail knowledge. (I also don't quite see why you think you want to do that, but I'll happily give you the benefit of the doubt and assume it does make sense for you.)

 

[gpsguy]

Just don't run (exit) the initial configuration wizard on bootup.

Even if you did, you could still boot it in the NAS and choose the "Reset to factory defaults" on the console setup menu.

 

[pirateghost]

Really it is no different than moving your install to a new computer which works fine.

 

[solarisguy]

@SmilingInSeattle, could you please share your experience? Wishing you all the best!

P.S.
I am personally interested, as with previous FreeNAS versions (9.2 and earlier), I occasionally needed a manual input from the console, when there was a different networking hardware.

 

[Bidule0hm]

As I've already said: there's no problem to install on one server and move the boot drive to another after the install. FreeBSD isn't Windows... :)

 

[SweetAndLow]

Should work fine other than the fact that usb3 isn't working correctly in freenas. Make sure to disable that in the bios and to do that you will need to use mouse and keyboard either physically or remotely.

 

[SmilingInSeattle]

Well, after fiddling several times with whether or not I could use a USB 3,0 port for either or both of the USB sticks, as both USB sticks were 3.0s (I concluded that, at least this time, the ZFS formatted boot disk must not support USB 3.0 drivers) I put the created and modified ZFS boot disk into my NAS, but was unable to ever ping the static address I had given it during the Initial Install procedure. I have to assume, at this time, the Initial Install failed to transfer because the Ethernet cards in the two machines were dissimilar.

I'm hoping a member of the FREENAS development team looks into this method of installation (since FREENAS encourages one to boot from a USB stick, has now provided ZFS mirroring, but has admit that USB technology is a "moving window" and actual storage space differs from manufacturer to manufacturer and model to model of the same quoted size. A quick alternative would be to allow the user to make the partition holding the zfs database to some size smaller than the "rest of the disk."

So, at this time I'll mark this thread SOLVED.

 

[solarisguy]

@SmilingInSeattle, you might have hit another buglet...

Were you able to go forward after deleting the network interface and then adding it? Option 1) on the console. Try that please!

 

[SmilingInSeattle]

OK, I'll be explicit.

My FreeNAS server is a Proliant N40L. Standard (suggested) procedure is to boot from a USB device and save all drive bays for one's RAID. After a recent 9.3 GUI snapshot upgrade, the server would not complete boot until started manually several times. After I finally got it up, I regressed one snapshot and then began learning how to mirror the boot pool. The USB used for booting was a 16GB Patriot chip (5 year warantee) Warantees are something I use as a subjective benchmark of how good the manufacturer thinks the device is.

When I tried to mirror the boot device, I used a Silicon Power 16GB USB thumb drive. It would not mirror. Upon investigation I found the Patriot thumb drive had a capacity of 15.1GB and the Silicon Power had a capacity of 14.8GB.

Since I had two of these same size Silicon Power thumb drives of the same size (and Silicon Power Warantees for life), I wanted to use one for the boot drive and mirror it.

I downloaded the ISO of the latest snapshot version of FreeNAS.9.3-STABLE that my server was running on to my Linux desktop, inserted the USB in a USB 3.0 port and dd'ed it on to the USB. I then shut down my desktop, inserted the boot ISO on the thumb drive and the unconfigured other thumb drive into various combination of USB 2.0 and 3.0 ports (see my earlier post). Anyways, I booted the desktop into the FreeNAS installer and told it to put the image of the installation on the other thumb drive, which it did. Then I Shutdown the desktop pulled both thumb drives, put the FreeNAS imaged thumb drive just created in the spot where the ISO thumb drive had been, and laid the ISO thumb drive on my desk. I booted the desktop into FreeNAS and went through an "Initial Install/Upgrade.

As my N40L has been assigned a static IP address, when I got to the 14 selections, I chose #1 and went through the procedure of assigning a static address at this point given in the FreeNAS documentation. (Exmple 3a. of the FreeNAS 9.3 documentation) The first time around I embellished it by adding Gateway and Routing information. Later I built a boot disk without adding the routing and gateway info as the documentation and on screen progress suggested the install had built-in "smarts" to handle a simple situation, as, this being, a home network is. for what it is worth, the static IP address used was 192.168.1.10 and Netmask was 255.255.255.0.

I then went to the headless N40L, which had been shutdown, pulled the existing boot USB and inserted the IP modified Silicon Power USB drive into the slot. From watching the leds on the N40L box I saw signs of booting, including disk spinup, but when I went back to the desktop (which had been booted back into Linux by then) and pinged 192.168.1.10, all I got back was "Host Unreachables" Each time I tried building a new install disk (and BTW, I did the same thing using a Samsung Laptop) I got the same results. What caught my eye as I wrote the post above suggesting the problem was solved was that the example shows an adapter called em0. The adapter I got when doing that step was er0. I could just be a misnomer, but it got me to thinking that the way the script runs, the IP configuration is for that particular type of adapter, and the adapter in my N40L is different from the adapter in my Linux desktop or in my Samsung Laptop.

My N40L has a Broadcom NetXtreme BCM5723 Gigabit Ethernet controller.
My Linux desktop runs a Realtek RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller
and my Samsung Laptop also rund a Realtek controller.

After this yesterday I ordered two 32GB USB sticks (kind of a waste) and will mirror the existing 16GB boot drive on to both of them. At least this way I'll have something I can boot off of if the 15GB drive fails again. I did look into seeing if I could find the same model of 16GB Patriot drive, but I couldn't. Anyway the mass storage market is too fluid to really use that approach for long, and store clerks (and probably the Chinese manufacturers) really would not be able to tell me the actual storage available on a generic sized thumb drive. That's why I suggested having the ability to control the partition size of the boot ZFS drive, kind of like the suggestion and reasoning behind temporarily turning off swap size when installing a replacement drive in an array until the drive successfully resilvers.

 

[solarisguy]

@SmilingInSeattle, I am sorry, I do not know what I was thinking..., as earlier I did read your first post about a monitor and a keyboard...

Did you try to use DHCP reservation instead of a static IP ? (You can later switch to static IP.)

P.S.
I mirror USBs right away when doing the initial installation (by selecting two devices).

 

[gpsguy]

Your stock Broadcom NIC on the N40L is a bge0.

If you had attached a monitor and keyboard to your N40L temporarily, you could have had this solved in about 5 minutes. Just do a reset, as I suggested above and/or let it grab a DHCP address (temporarily). Then restore your configuration.

 

[SmilingInSeattle]

@solarisguy Just reread your post.

I am now going to try to the delete the network interface and install. I'd tried that before, but I'll see if I can delete the adapter and figure a way to get it to probe again for the adapter

 

[pirateghost]

When using another system to install the OS onto USB, don't configure any settings after the install. Just boot it on your final server and use the dhcp address it gets assigned by default.
Set the static after you get to the web GUI

 

[SmilingInSeattle]

@gpsguy That option was always available. I've done that before and hated lugging all that hardware to the server. I have to pull the server out into the room which means recabling as I've got more drives than one box holds. My monitors are all in use and the smallest one is 24 inches. I've been looking into 7-inch touch screen monitor kits, but haven't found one with the specs I want at the price it's worth. Anyway, this is and was an experiment!

 

[gpsguy]

Back in the day, you could have purchased a Microserver Remote Access Card which would have given you IPMI access to it. The cards are discontinued, so if you wanted one now, you'd have to look on the used marketplace.

 

[SmilingInSeattle]

OK, the idea of deleting the network interface and reentering everything didn't work. I did that just now, rebooted and gave it 15 minutes to come up on the network. No go. I'll try the dhcp approach, but may run into a snag as to the way I configure my router. (I have only allocated a small window of the subnet to dhcp, and in that window I assign addresses based on mac addresses of the equipment I own.