EMMC on SG units
-
What would you like to use it for?
I rather expected pfSense itself (OS and the like) to be on the flash with the SSD being used for data.
-
Running some kind of split flash/hard-drive install gets complex. There are many combinations you might want. People have done it using Nano to boot from flash and running a harddrive for logging and Squid cache say but it's becomes full custom at that point. It won't survive a firmware update.
Steve
-
In the current config on the 2015 Mustang with a 5.0-liter V8, the passenger seat appears to be unused. Are there plans to make use of this empty space in the future? :P
-
In the current config on the 2015 Mustang with a 5.0-liter V8, the passenger seat appears to be unused. Are there plans to make use of this empty space in the future? :P
Only if you skipped the Recaro seats. They're pointless unless you're driving, and even then, they're too uncomfortable to sit in for long.
-
I'd say it's more like leaving the glove box empty, not a seat. :-)
pfSense in general has not been inclined to use multiple storage sources without manual fiddling. It's not too difficult to gpart/newfs a disk and drop it in fstab or a shellcmd to mount. But to support that officially is a lot more work.
It would make a nice place to drop backups or other infrequently-updated data.
-
pfSense in general has not been inclined to use multiple storage sources without manual fiddling. It's not too difficult to gpart/newfs a disk and drop it in fstab or a shellcmd to mount. But to support that officially is a lot more work.
Understood. Thanks.
-
I'd say it's more like leaving the glove box empty, not a seat. :-)
I'll admit it wasn't a perfect analogy. :P
-
I'd say it's more like leaving the glove box empty, not a seat. :-)
pfSense in general has not been inclined to use multiple storage sources without manual fiddling. It's not too difficult to gpart/newfs a disk and drop it in fstab or a shellcmd to mount. But to support that officially is a lot more work.
Actually it's more like a glove box that's been welded shut. It's there, it's paid for, but it can't be used if there's another storage device in the unit without some customization.
At some point in the product design there was a decision (implicit or explicit) not to support multiple storage devices in a single unit. It would have been easy if designed in from the project start. Current SG units have 4GB eMMC components which are fairly low cost so it's not really a big issue to leave it idle.
It may be a different matter with the upcoming SG-8860 which has 64GB eMMC. Users of that unit whose needs eventually exceed the size of the (admittedly large) eMMC would be idling a component that is ~ 25% of the unit cost by my 2-minute estimate.
-
Did you ever thought about, that this nand storage will be there for setting up pfSense on
and then another storage as a mSATA could be used as a cache or alternate storage option?Current SG units have 4GB eMMC components which are fairly low cost so it's not really a big issue to leave it idle.
Would it better to go with a changeable storage such as a CFCard, CFast Card, SATA_DOM, IDE_DOM,
USB Stick or whatever?It may be a different matter with the upcoming SG-8860 which has 64GB eMMC.
Users of that unit whose needs eventually exceed the size of the (admittedly large) eMMC
would be idling a component that is ~ 25% of the unit cost by my 2-minute estimate.Why? You can install on the nand storage read only and using another storage as for caching or
backups also. -
@BlueKobold
@BlueKobold:
Didi you ever thought about, that this nand storage will be there for setting up pfSense on
and then another storage as a mSATA could be used as a cache or alternate storage option?That's exactly what I wanted to do. I was a little disappointed reading this thread.
@BlueKobold:
Current SG units have 4GB eMMC components which are fairly low cost so it's not really a big issue to leave it idle.
Would it better to go with a changeable storage such as a CFCard, CFast Card, SATA_DOM, IDE_DOM,
USB Stick or whatever?Hmmm, while a USB stick would work easily, I wasn't aware that the SG series supported the other options without some non-trivial work. From my playing around it appears that the boot priority on an SG series device is
USB ports
mSata
eMMCI haven't tried booting from the miniPCIe cards. You can change the primary boot device at boot time through the serial port, but that not very convenient (I so wish the SG series had a BMC for the IPMI BootDev command). But I imagine you would be able to update the boot blocks on the primary boot device to load the OS and applications from another device on each boot. I haven't worked with BIOS for some time (almost everything I work on is UEFI these days) so I may be out of date.
@BlueKobold:
Why? You can install on the nand storage read only and using another storage as for caching or
backups also.May I suggest refer you to these comments from earlier in the thread:
Running some kind of split flash/hard-drive install gets complex. There are many combinations you might want. People have done it using Nano to boot from flash and running a harddrive for logging and Squid cache say but it's becomes full custom at that point. It won't survive a firmware update.
Steve
pfSense in general has not been inclined to use multiple storage sources without manual fiddling. It's not too difficult to gpart/newfs a disk and drop it in fstab or a shellcmd to mount. But to support that officially is a lot more work.
These comments suggest that you can't use the multiple storage devices with some customization, as I suggested. The issue is whether you are willing to redo your customization after each update. If you are, great, more power to you. Or you could forego updates, but that's not a super great idea. Or you can rebuild with your changes each time you update, if your changes are extensive.
There's all sorts of options, but what you can't do (if I'm reading stephenw10's comment correctly) is make the changes that you suggest and have them promulgate to future versions automatically.
-
I have also thought about doing this in the past - putting a disk or SSD in a nanoBSD system and using it for /var and /tmp. Now that there is code for selecting where these partitions live it might be just a bit easier to do the run-time code. The config can remember where they should be and mount them (and/or generate fstab and/or…).and let you
The part that is more work is making GUI and/or console menu code that will display the disk-style devices available and free and let the user choose how to partition them and what parts of the file system to put where. There could be a lot of possibilities as people put different devices in their systems on different connections/buses - old IDE, SATA, eSATA, USB-sticks, USB-disks... -
As a good example of what is required to do this currently look at this:
https://forum.pfsense.org/index.php?topic=67823.0There are various things you can do to move most (perhaps all) of that within the config file. Using the Shellcmd and Filer packages for example. However if will still fail at first boot after an upgrade.
Steve
-
putting a disk or SSD in a nanoBSD system and using it for /var and /tmp.
Exactly this I was meant!
