i recently purchased this board to build a NAS with 10G and some spare network ports to replace the home network switch as well.
My hardware now consists of
– A Mini-PCIe riser card to an x1 edge connector from ebay housing a NVME SSD for performance storage
– A Mini-PCIe 2-port SATA controller (ASM1062 based), with two disks in RAID1 for archiving
– Finisar/intel 10G optical SFP+ i had lying around (ftlx8571d3bcv-it)
– 8GB DDR4/ECC
– As wireless “bmc” (power on/off, UART) and WiFi interface i’m using a spare VoCore module.
– a “widenend” case of an old qnap NAS 😉
– Add 3rd PCIe expander for another 10Gport (need to change the bottom/WAN slot from USB to PCIe)
– Switch off SFP+ transmitter when unused.
– Fanless cooling to have a living-room NAS
For the 2nd pcie slot to operate in PCIe rather than SATA mode got a patch from solidrun for the u-boot/linux dtb, see https://github.com/fesc2000/sr-4080 (i’m going to put stuff there i might adapt/develop for this project going forward, maybe this draws some attention).
Software is Debian image frol SolidRun with 5.4.0-0.bpo.3-arm64 kernel, management interface is webmin.
So far this works quite nicely, but there are some topics i hoped to find a community to discuss with. The forum on the solidrun site seems dead, in don’t even get a login from them. Here i was “surprised”, to phrase it nicely, to get a http registration page by default, and the https certificate is expired .. I’m not sure what to think about this level of “professionality” and wonder where my password will end up. Not to speak of the “unrelated” posts in this forum.)
Nevertheless, some observations and questions:
– Is there any other active forum/community i did not find?
– The power consumption in idle is a little bit high for me (like 25W). I’v seen some idle mode patches, are they worth trying? I don’t want to trade power with stability issues.
– At the moment, i got an occasional (ca once/2days) kernel/Core freeze (rcu_sched detected stalls on CPUs/tasks) for no obvious reason. Has anyone observed such thing anywhere else? Temperature is OK.
– This might be memory related, but i’m not sure if proper ECC detection is enabled in the kernel, like reporting single-bit corrections which could indicate a shaky memory interface (at least i haven’t seen any messges from the kernel).
– Anyone knows if thee is an easy way to enable a watchdog in case this happens? At the moment i’m doing a power cycle via my bmc (need to change to reset).
– I’ve written a small PWM fan control tool (see my repo). Is there some kernel module doing this in a standard way?
– The 4-port switch on the 2.5G interface seems sensitive. At startup there are a bunch of “switch 0x3400 detected: Marvell 88E6141, revision 0” events before it settles. It was even worse (basically endless messages) when i used a weaker PSU or loaded the onboard 5V more to power a SATA drive. Now that i use an (over-dimensioned) PC PSU it’s better. In general, all power rails looked good though, maybe it’s the reset signal ..
– The 10G interface needs some time (seconds) to settle/lock when coming up (lots of CRC errors, but then it’s stable)
Last but not least, i consider testing/switching to OpenWRT (although i don’t need any fancy network stuff, i put all network ports into a bridge). Any hints/suggestions which one is maintained better? Especially in terms of network performance, on the 10G port i get ca. 6Gb/sec rx and 4Gb/sec tx (4 parallel streams, no jumbo frames) at best with iperf3, which could be better for a SoC with acceleration hardware.
Also i had hoped that the pcie/nvme performance would be better, even with only one PCIe lane, ~6GBit read should be possible. The “network backup performance” towards the nvme via the 10g link reaches up to 2.5Gbit/sec, which is fine.
(hope this is not off-topic, but, as mentioned, it seems there is no such thing on this forum 😉