PFsense on Alix board : speed of the Compact Flash card ?



  • I'm in the process of replacing my current CF card installed in my Alix based pFsense.
    But, finally, is CF card speed so important ?
    What are the CF card speed exact requirements ?

    Thanks for your inputs !



  • Yes and no.

    Yes, because in most cases with NAND, "Speed" relates to quality; faster flash chips tend to be (but aren't always) more reliable.  A very high quality card could be used for a full install without issues rather than a NanoBSD.  I have a couple systems which have been running full installs on SLC and MLC cards (most cards are cheap TLC) for 4+ years.

    No, because that board is so slow you won't notice the difference between a high-end card like a Sandisk Extreme Pro and a "Standard" card like the Sandisk Ultra.

    Just stick to a name brand and you'll be fine.



  • I think NanoBSD is there especially for reliability. The whole point of having a NanoBSD install is to have the system running from a read-only storage medium, and no writing back never (except for the rare case of configuration saving process). This makes it run reliably almost forever, at least trying to minimize defects coming from storage level at the minimum possible.

    Speed doesn't matter here, where reliability is #1 factor.

    Speed of writing can be an issue if logging back to the card - but I'd definitely avoid that. Why log to the card? Log to the RAM, a pfSense box doesn't need to be restarted. Or if you definitely need to archive all the logs, set up a syslog service in your network and log there…



  • @robi:

    I think NanoBSD is there especially for reliability. The whole point of having a NanoBSD install is to have the system running from a read-only storage medium, and no writing back never (except for the rare case of configuration saving process). This makes it run reliably almost forever, at least trying to minimize defects coming from storage level at the minimum possible.

    Speed doesn't matter here, where reliability is #1 factor.

    Speed of writing can be an issue if logging back to the card - but I'd definitely avoid that. Why log to the card? Log to the RAM, a pfSense box doesn't need to be restarted. Or if you definitely need to archive all the logs, set up a syslog service in your network and log there…

    A read-only root file system doesn't make NanoBSD more reliable.  Bad blocks still occur, especially on the cheap TLC NAND used in many CF & SD cards, and hitting one can cause your system to lag or panic, depending on what it was and when it happens.



  • And…I think pFsense still uses the CF card to store his default logging ?
    Or am I wrong ?


  • Netgate Administrator

    More importantly most of the 'fast' CF cards, 300x etc, are achieving those speeds through UDMA transfer modes. In the standard NanoBSD images DMA is disabled so you won't see anything like that performance. It might help you write the image to the card faster.  ;)

    Steve



  • @chercheur:

    And…I think pFsense still uses the CF card to store his default logging ?
    Or am I wrong ?

    Logs, RRD data, DHCP leases, all that sort of changing data is stored on /var/ and /tmp memory file systems in RAM.
    You can set the system to write RRD and/or DHCP lease data to the CF card at the interval you want (or never). On clean shutdown RRD and DHCP data is written to CF so it can be recovered during the next boot. If you suddenly remove the power, then that data is gone - you will have just the RRD and DHCP that was last saved to CF.
    Logs start again at each boot. If you want to save log records, then log to a syslog server.