[Fixed] 2.1.4 Nano update files are all VGA version

pvoigt

Thanks, Steve, for pointing to a possible interchange of vga and non-vga (e.g. pure serial) images on https://updates.pfsense.org/_updaters/.

It will help to avoid trouble for users having serial access only to their pfSense boxes. I have an Alix and a Soekris box and have decided to wait until the situation is more clear.

I agree with you from the file names in the e.g. in the md5 signature files. But in this case I would expect equal checksums:


# cat latest-nanobsd-2g.img.gz.md5
MD5 (pfSense-2.1.4-RELEASE-2g-i386-nanobsd-vga-upgrade.img.gz) = de5d516d3b1bd27caceeb1b546e78395
# cat latest-nanobsd-vga-2g.img.gz.md5
MD5 (pfSense-2.1.3-RELEASE-2g-i386-nanobsd-vga-upgrade.img.gz) = 597c26902fa1a91af3d26ffa6a94f379

As checksums are not equal things might even be more complicated.

Regards,
Peter

EDIT: Hhm, have just seen that the vga images even has an old file name.

stephenw10

You are comparing 2.1.4 with 2.1.3. ;)
The two 2.1.4 update files do have the same checksum.

Steve

pvoigt

@stephenw10:

You are comparing 2.1.4 with 2.1.3. ;)
The two 2.1.4 update files do have the same checksum.

Steve

Hhm, but which files should I have compared?

Right now https://updates.pfsense.org/_updaters/ has been updated. Time stamps changed from May to June and non-vga versions are now more recent:


# cat latest-nanobsd-vga-2g.img.gz.md5
MD5 (latest-nanobsd-vga-2g.img.gz) = de5d516d3b1bd27caceeb1b546e78395

Now MD5 sums are identical and file name has changed and your thesis is confirmed :)

Did I compare wrong files? Am I the only one seeing this update on https://updates.pfsense.org/_updaters/?

pvoigt

Well, I have just successfully updated an Alix from pfSense 2.1.3 to 2.1.4 using "Manual Update" method from the webGUI. And in particular there were no issues at all with the serial console :).

I have used the NYI.net mirror and this image file: pfSense-2.1.4-RELEASE-2g-i386-nanobsd-upgrade.img.gz.

This again confirms your suggestion, Steve: The non-vga images files on https://updates.pfsense.org/_updaters/ are obviously wrong.

After this success I am going to update my Soekris as well with the same image file.

Regards,
Peter

jporter

We're looking into this, there was a change in the file namings that may have confused the uploading scripts.

stephenw10

Thanks for the update. :)
I'm suprised that there aren't more posts on the forum about this, it must be affecting fewer users than I thought.

Steve

charliem

@stephenw10:

Thanks for the update. :)
I'm suprised that there aren't more posts on the forum about this, it must be affecting fewer users than I thought.

Or we just have faith that you will persist until it gets fixed :)

richardstubbs

CONFIRMED – Console upgrades on ALIX boards will currently cause the loss of the console on reboot as the “auto update from a URL” option will cause the VGA build to be installed.

ALIX 2D13 4GB CF Model "update from URL" -> "Auto" (https://updates.pfsense.org/_updaters/latest-nanobsd-4g.img.gz) -> get's "pfSense-2.1.4-RELEASE-4g-i386-nanobsd-vga-upgrade-20140620-1259.img" installed and loss of console.

stephenw10

Given that and the number of Alix boxes out there I expected a flood of complaints but I see hardly any, only here and the X700 thread below. :-\ Are the vast majority of Alix users running 512MB images which havne't been updated? Perhaps most are using Netgate sourced boxes which point at a different URL?
This is obviously just a minor scripting problem and I'm sure it will be quickly fixed I just find it interesting that it hasn't affected more people. As an asside it would great to get a breakdown of how the known pfSense installs are distributed across platform type and 32bit/64bit if such numbers are known.

Steve

-flo- 0

Given that and the number of Alix boxes out there I expected a flood of complaints but I see hardly any […]

I use to sanity check before I apply any updates. This thread here is a good example why this makes sense. Although I can only speak for myself I assume that there are more people out there acting this way.

As an asside it would great to get a breakdown of how the known pfSense installs are distributed across platform type […]

I agree. Perhaps a breakdown of user's ages would also be interesting. ;D The older I get the less fun and the more work update seem to be. :D

richardstubbs

I suspect they just simply have not noticed yet (the VGA image will work fine, just no console as most people would upgrade from the web UI (and can fix it this way by pointing the update URL to a correct appropriate image on a mirror) they haven't discovered the lack of console yet)

We drop to console on initial build (to check its working) and initial interface assignment, but I just happen to be shipping out some preconfigured D213's (the xml configs come from our version control system) that were built last week and didn't want to ship them out with 2.1.3 and then have to raise tickets for a scheduled update for 2.1.4. Consoled in, switched the WAN to DHCP and did an upgrade then would have redeployed our configs via the console, however as we have discovered no console on reboot

I rolled back to the other working slice and upgraded from a good mirror image and reapplied the XML configs.

I did initially think that update URL / naming convention had changed in the config but this was double that this was not present in the config

stephenw10

Good point. The X700 users who orginally asked questions reported their unit failed to boot so I assumed the lack of video hardware was an issue. Thus I expected the Alix to similarly fail to boot but clearly that isn't the case.

The decision to upgrade my own home box has been removed since Microsoft just took down my dynamic DNS! >:(

Steve

tix

For what it's worth, the update broke my x700 also…

I still had the serial console at 115200 but it seemed to be using a different boot loader - reminded me of GRUB on Linux. The default boot option would load the DMA module and my CF card did not like that. I was able to recover by selecting the failsafe option (#4 on the menu) and pfSense would boot normally. Once it was up I could copy my good 2.1.3 slice over the update and be operational.

I'm waiting and watching for confirmation that the update servers are fixed before attempting again.

stephenw10

Completely forgot about that. Yes, the reasson the vga images won't boot is that they have DMA enabled and the X700 CF slot doesn't support it. However I thought that was true of the Alix also. :-\

Steve

eweri

Same here for me: upgraded via serial console after reboot the console is gone.

Who do I have to contact to change this?

Bye,
eweri

corvey

Any updates on when the next release will be out to fix the auto invoker update?

This version is broken and can not boot unless I manually hit the number 2 through serial:

image share

shenion

I fixed mine by using the updates manually from:

http://files.bgn.pfsense.org/mirror/updates/

The web gui update just timed out, I had to use the console.

Now the console works. Seem part of the boot is still at 9600 though but at least there is a console now.

Note using the nanobsd files from the default location instead of the vga one the updater switched to completely disables the console (no longer in /etc/ttys.)

It looks like they are moving to one version for both.

Guest

I have some boxes headless with nanobsd 4G and I'm somewahat confused now… should I try to update them from the GUI or will it break the console? Why is there no official response to this apparent issue? Is a fresh install without risk?

Any help highly appreciated... ???

stephenw10

So if you update your box and end up with the nano_vga image you will loose the serial console. However that may not be a disaster, you can always switch back to the previous slice if you have to. A bigger issue is that the standard nano images have ATA DMA disabled by default in the loader.conf file. This is because many older CF slots did not have the required connections to support DMA such that if you insert a newer DMA capable card you just get errors. If you box has such a slot and your CF card reports it has DMA the kernel will try to use it and will fail to boot. That is much more of an issue since you can only then recover by choosing the alternative boot slice at the boot loader via the serial console or by re-writing the image on the card.

There were a few comments some time ago that suggested it may be possible to have both vga and serial consoles on the same image, initially I thought that may be the case here. Unfortunately this coinsided with my being away so I have no way of testing anything.

Steve

bennyc

@jporter:

We're looking into this, there was a change in the file namings that may have confused the uploading scripts.

Awfully quiet here… no progress?