grep

Returned zram swap back. Rebooted. And once again:

In short: disabling zram-config.service and using "pure partition swap via ssd" causes problems HC-1, armbianmonitor -u = http://sprunge.us/aRZc UPDATE: "New year problems party" continues: armbianmonitor -u = http://sprunge.us/ZYLE

Has anyone tried to use a fan with Odroid HC-1 ? Too hot for my default load scenario (75..80 deg. cel.), even had to make an "ugly rack" (with hdd location below HC-1) to avoid hdd overheat and allow free airflow from heatsink. There are 2 "dot"-connectors marked "GND" and "12V"; looks like they are for external cooler, am i right ?

@Igor It's meaningless, because my "black cube PSU" provides OVERVOLTAGE by default (see image attached) So, i'm just an "extremely lucky" one who got 2 unstable PSUs. Absolutely awesome. Will wait for 'reference PSU'. @tkaiser i'll say it: ok, looks like you were right So: most likely, the set of sporadic problems was caused by usage of 2(!) unstable (OVERVOLTAGED) PSU's one-by-one.

grep@orangepiprime:/home$ sudo btrfs scrub status /home scrub status for 835d7f1d-2047-47b1-9dcc-24bbc3be48c0 scrub started at Sun Dec 17 14:49:30 2017 and finished after 00:42:45 total bytes scrubbed: 94.53GiB with 0 errors grep@orangepiprime:/home$ df -h /home/ Filesystem Size Used Avail Use% Mounted on /dev/sda1 430G 96G 333G 23% /home stress --cpu 8 --io 4 --vm 2 --vm-bytes 128M --timeout 501s stress: info: [21289] dispatching hogs: 8 cpu, 4 io, 2 vm, 0 hdd stress: info: [21289] successful run completed in 501s See no problem too (BTW, your SoC is really hot under stress, so my previous guess about SoC overheat looke more and more like wrong one): But i will replace PSU with "black cube" and will "just watch if something strange happen". After that, seems like the best i can do is to wait for "reference PSU" from Xunlong (but HC-1 will arrive earlier some maybe i just will forget OPi Prime as a nightmare).

@tkaiser @chwe @Igor @TonyMac32 Sorry for interrupt, but this one does not look for me as the same PSU problem (i turned off all except monerod, so no CPU/IO load at all; http://sprunge.us/RGLj ). Should i start new topic or log is enough for you to investigate it "one day"?

If we are talking about uptime, 60+ is not too bad (with ubuntu), so let it be with HC1. But i do prefer something kinda # uptime 00:38:34 up 1368 days, 5:20, 1 user, load average: 0,11, 0,09, 0,09 even with cost of "lack of maintenance" if security allows that.

100% agreed, nothing to argue with; @tkaiser even purposed to mesaure voltage via GPIO ... but i don't have enough internal motivation to perform tests once again + i definitely need to "define the correct questions/hypothesises" + "determine correct methods to find the proves/discards for those questions/hypothesises" ... yak! too much work for such "question": it's much easier to buy "reference PSU" That's a good approach when you have to make big deal with lack of resources/small team. See nothing unusual with it, even more, i was really astonished with a lots of posts of @tkaiser one-by-one in this thread. Separately: @tkaiser, thank you for participation, i really appreciate that. BTW: does Armbian project need good photos of OPi Prime PCB ? (you've mentioned it in one of your replies) I can make it, but with my radiators in near day, or without them after my Odroid HC-1 will arrive. Looks like it will go there, as @TonyMac32 had adviced before: it's really too hot to use it without extreme caution. Oh, please, no more fun like that! Better wish me something kinda "Stable powering and smooth working"

So, something like conclusion: 1. You should never use Y-USB cable for powering USB-HDD: it could cause dangerous side-effects like powering your board (and you will not even know about it); 2. Real reason for MULTIPLE "translation fault" and "usb device reset" is still unknown to me: i used two different PSUs and got errors with both, but real experiments (voltage measurement under stress) was with only one, and, finally stabilized the behavior (the reason found was Y-USB powering for HDD, but what were the reasons for all the previous crashes ? ...) 3. By "sum of facts" i could suppose 2 things: - There is an overheat of SoC in "electric closet" which causes strange effects (weakly confirmed, need more different experiments to prove/discard) - PSU "lied on back during tests" (this is "normal position") while in "near by electric closet" it "lied on left-side/hanged on wires" (... yeah, i know it sounds insanely, but who knows whats happening with "hot airflow" within this veeeeery hot PSU when it "lies on left-side" and there is no free airflow to the outside) +funniest thing: never thought that board could be powered via USB, always thought that there is a "current in-flow protection from outside" (kinda "something diod-based"), but looks like "it depends on PCB-implementation" and for Orange PiPrime answer is "yes, you can continue to power the board via USB if you already powered it up"

@tkaiser so, in fact, by using (if generalize widely) "cheating methods" i got a lot of support from you using your "feeling of responsibility for the Armbian project". Do no offend please, that definitely was not my intention (i would performed all those tests anyway because of my nature, but it was good experience: to ask for help and to share results with others + learn a couple of lessons kinda "never use Y-USB" and "Check the PSU before you going to replace it") And i think i will replace my OPi Prime with Odroid HC1 (which i hope won't be so problematic)

Walls are spoiled, except the last one (i'll spoil it after completion).

@tkaiser Mmm.. cant understand this; as i understand "almost nothing" had changed. Do you remember this: Now it's just a: PSU screw terminal <=> two thick wires taken from "standard black computer PSU power connector internals" <=> twist, covered with insulating tape + multimeter connected to this point <=> two a bit thiner (but thick enough) wires ended with 1.7mm jack (cut from 3A PSU). So, there is only one thing changed: i removed separate power supply via Y-USB-CABLE @tkaiser i've attached another foto, wires connection is pretty clear from it

STEP3 candidate. After a while, and, of course, STEP2 completion.

It was not about "savings", i looked something different from "standard black cubes" with 5V/3A, and found that one. They have a wide range of PSU's so i thougt that they know what the do. In fact, indeed: for example, should not there be any protection on board from such current flow ? +there was a recommendation from other user ( @guidol ) to use separate power source for USB-HDD, so this is definitely "not obvious" (at least for me and him) Already ordered one from Germany with 4A PSU. There is no official shipment to Russia, but there are a lot of ways to get it. Right after STEP2 (see my updates). And looks like (at least for now, after 1h:10m of testing) the problem is not with PSU (the question is "how long shoud it run in critical mode to we make sure that it is stable enough") Disk drive is SSD, so no spinning. Sounds strange: you deny lessons learned from negative experience. What is 'proper' or 'not proper' ? Saying 'not proper' you're making a lot of non-proved assumptions (for example about PSU). And about 'expirienced' users: once again, what will you and @tkaiser and @TonyMac32 would say if this "shitty psu" will remain stable in next 24 hours under stressed mode ? The question should be set in different form: "do we have enough proves of our assumption" but not in form "my expirience tells me that problem is with PSU, so, because i'm well-expirienced, trust me, you're wasting your time".

PART2 This time i used USB 3.0 "single" cable and connected multimeter to the place of twist of wires from screw terminal & 1.7mm jack (you could see all the construction on photo, i've moved all the parts from electric closet to the table, this allows to monitor voltage continuously) I've skipped tests @1104 MHz (see no point), so: 1. No load / 1152Mhz: 5.11v (funny, looks like limited electric closet space without normal airflow gives ~ +10 deg. celcius of SoC temp on idle) 2. 1 minute after start a lot of IO/CPU intensive tasks: flapping 5.08..5.10V 3. 10 minutes after start a lot of IO/CPU intensive tasks: flapping 5.08..5.10V (i'm really impressed with temperature: max is 35 deg, average is ~30; defenitely electric closet was a bad idea to place the board into) 4. 20 minutes after start a lot of IO/CPU intensive tasks: still flapping 5.08..5.10V (temperature is still awesome ~28, but PSU became really hot) 5. 30 minutes after start a lot of IO/CPU intensive tasks: still flapping 5.08..5.10V (temperature is ~28, PSU is still really hot) 6. 40 minutes after start a lot of IO/CPU intensive tasks: still flapping 5.08..5.10V (temperature is ~30, PSU is still really hot; looks like i should wait PSU to overheat and shot start behave itself unpredictably) 7. 50 minutes after start a lot of IO/CPU intensive tasks: still flapping 5.08..5.10V (temperature is ~30, PSU is still really hot) 8. 2h40m after run ... and nothing happened: 5.09 is stable now, temperature is stable ~30.

Daamn .... unplugged the 1.7mm-jacked cable to make PART2-experiment with "standard cube psu", got cable in my hand and shocked ... THE SYSTEM IS STILL ONLINE (fan is working, network led is blinking)! All the time it had some kind of DOUBLE-powering (USB charger via Y-USB-CABLE + 1.7mm jack from PSU). How is that possible at all ? @tkaiser you was right, if the Y-USB-cabling is prohibited by USB specification than that is because of "fear of side electrical effects" that partially described in https://superuser.com/a/1022965 @guidol that's a lesson for both of us: i think starting from this day you definitely should not recommend to use Y-USB-CABLE (i.e. separate power for USB) to anyone, side-effects are absolutely unpredictable Take a look at the photo attached: - left one: HDD is powered via Y-USB-cable, 1.7mm jack is out, board is not powered - in the middle: both 1.7mm & Y-USB-cable attached, board is powered on - right one: 1.7mm jack is out, but the board is still powered on, i.e. "powering the board through Y-USB-CABLE" Looks like PART1 needs to be revised with "normal USB cable" and the same PSU.

@TonyMac32 you know, the problem is that the first time i used another 5V/3A PSU ("standard black cube") but got the same symptoms multiple times ("usb device reset" and "translation fault"). Looks like it's time for PART2 with "black cube" .. (but in fact, this "dance with dragons" is a bit annoying: i just want it to work, but absolutely do not want to fight against "the army of undead boards and PSUs")

PART1 (hope a will have enough patience to write other parts) HDD has additional "power" connection to USB-charger (via Y-USB-cable) All measurements made on screw terminals of PSU with multimeter (image attached) TL;DR: PSU becomes unstable after a while on heavy load scenarios according to measurement on screw terminals: after 40 minutes of work voltage jumps to 20V, i.e. we have OVERVOLTAGE and after that "translation fault" and "USB device reset" ( http://sprunge.us/FKMO ) remark: i can't measure "all the time", because the PSU and board are located inside(board) and outside (PSU) "electric closet", so i had to "run to PSU, measure, and run back to computer". @tkaiser @Igor @zador.blood.stained @guidol have you ever seen something like that ? 1. Default, no load: 5.14V 2. 10 min after 1104MHz: 5.17V 3. 20 min after 1104Mhz: 5.15V 4. After stop all the processes for a couple of minutes / 1104MHz: 5.14V 5. "Uncontrollable hell begin": Lots of IO and CPU-heavy processes (but CPU-overload causes drops in IO) @1152 MHz: 5.12V 6. 10 min after "Uncontrollable hell begin" @1152MHz: 5.12V 7. 20 min after "Uncontrollable hell begin" @1152MHz: Flapping 5.11..5.12V (measurement took ~ 40 sec) 8. 30 min after "Uncontrollable hell begin" @1152MHz: Still flapping 5.11..5.12V (measurement took ~ 1 minute) 9. 40 min after "Uncontrollable hell begin" @1152MHz: INSANITY 6.2..8.7V (measurement took ~ 30 sec) 10. 42 min after "Uncontrollable hell begin" @1152MHz: ABSOLUTE MADNESS 16.1..19.7V (measurement took ~ 30 sec) 11. 45 min after "Uncontrollable hell begin" @1152MHz: LOWERING 12.1..12.5V (measurement took ~ 30 sec) 12. ~50 min afer "Uncontrollable hell begin" @1152 MHZ: 5.13V and everything is dead: http://sprunge.us/FKMO

Absolutely forgot about adjustment (don't even know the default adjustment level). Need to take a look, but not near my board now; thanks @tkaiser

SSD Samsung 850 Pro 512 Gb (...please don't remind me about "hunting squirrel with a howitzer"). Did no try to search deep, but official specification (https://www.samsung.com/us/system/consumer/product/mz/7k/e2/mz7ke256bw/850PRO.pdf) says (for "active" state) V = "5", Power = 3.5W (5Vx0.75A, as i assume); ok, let them lie and it's 7.5W under load (5Vx1.5A); i have a USB-charger with 2.1A (let it lie too, and there are only 1.7-1.8A which is enough). BUT: i also do have a SATA2USB with"unknown power characteristics" (so, there are at least 2 points to measure Voltage/Amp: SATA OUT to SSD, USB-IN to SATA2USB) and first one is "not the easiest thing to measure" PSU screw terminal <=> two thick wires taken from "standard black computer PSU power connector internals" <=> twist, covered with insulating tape <=> two a bit thiner (but thick enough) wires ended with 1.7mm jack (cut from 3A PSU). All wires are thick enough IMHO, but i agreed, that real voltage/amp level is a subject to check (potential losses / additional resistance in points of twist i hope are negligible because of "contact surface length" which is formed as ~10mm of "naked" wire from each side ) I will try to measure voltage and amperage under load, but it will take a while. P.S. +I also ordered Odroid-HDR-HC1 (because i just want things to work) so, maybe OPiPrime will be added to my collection of "things that just does not work out-of-the-box as needed for me"

@tkaiser Thanks, a lot of info, need to think about all you said.

Then move it to correct subforum, see no problem. This is just your POW, nothing more, agreed ? I really should "hide" those walls with quoting, but in fact, there was at least two ideas based on output: about external power supply for USB-HDD and "cpu underpowering => test with lower freq". This is "more then no support", so 1) thanks for all 2) walls are working. And that's my opinion, and, as i guess, it does not contraversal to forum usage rules.

I've already got it So, i try to do my best to solve my problems by myself with help of volunteered people around.

ok, took latest with this commit (strange, but see no changes in this commit except "if-ing" & more strict checking to detect absence of temp-sensors, but doest not matter) Anyway, crash with 1104MHz ( http://sprunge.us/cacb ). Reasonable to assume problems with my USB external case (thing i could test by myself)?... Will check in a couple of days. @Igor So, unleashing 1152MHz (instead of previous max of 1108Mhz) does not look successfull yet. At least with my board +more: with 1104MHz & current load (monerod & bitcoind with cpulimit 200+150) i could not even perform a login via ssh until it crashed ("key exchange timeout"), so, i suppose, it was a "real-life stress-test".

@Naguissa Max temp under load is not more then 50, usual range is ~44..48. Not too much IMHO, but need to see CPU specification (but lazy a bit ... ) @Igor i've set 1.1Ghz via "userspace" governor to evaluate stability with low freq (and, therefore, voltage). And, BTW, http://sprunge.us/ still returns 500, for example: http://sprunge.us/hUZM (i don't know is that yours or external service, but just informing)