DS918+ & Qnap UC51T keeps disconnecting #58
Comments
|
I am having the same issue with a nearly identical configuration - the only difference from @T13nou is that I'm running DSM 6.2. The driver works well for a while and then fails, requiring me to either reboot the NAS or restart the driver package if I'm able to connect to the NAS via a different NIC (sometimes I can, sometimes not). |
|
Same here. Ds620Slim and this QNAP adapter, DSM 7. Trying to resolve the issue, I have installed additional USB fan blowing on adapter and switch (also very hot). Now they are both ice cold, so temperature is not the cause here. |
|
Ditto here on a DS920+; i'm now rebooting every monday as a precaution.. It starts to fail after about a week. |
|
My DS918+ running DSM 7.0 via a Sonnet Solo5G also drops the USB Ethernet connection once every two weeks or so. I generally shut down and restart the driver package and its all good to go. I've not been able to determine a cause so far but I only use this connection for moving things around from one computer. I use the onboard ethernet for all other devices on my network to connect to so it does not get much traffic. BTW I've not had much luck using the front USB (very slow speeds) so always use the rear USB port. |
|
@sparkplug37 that's interesting, I'll move mine to the rear and see if there is an improvement. I've been using the front as I believe that's what @bb-qq recommended? Maybe I'm wrong. How is everyone handling failover? I tried to do an active/passive link aggregation with the onboard NIC, but it didn't work (no errors were thrown, but the interface for the 5G adapter completely disappeared until I removed the agg). I'm trying to find a way to have It failover to the 1G NIC if the adapter/driver fails because I need close to 24/7 uptime. I'm using the QNAP adapter. |
|
I tested both front and rear USB and got about .5gbe on the front usb and over 2gbe front the rear tested with iperf3 over a 10gbe switch. I also liked the rear fans cooling the adapter as an added bonus. I have no failover as I just notice it’s down and use the onboard Ethernet to get back in. It seems 1 in 8 failures means a driver restart is not enough so a box reboot gets it back. |
|
So my ethernet adaptor stopped again this morning and I caught this call trace output from dmesg along with many 'aqc111 2-2:1.0 eth2: kevent 2 may have been dropped' messages. I do not know if its helpful to anyone much more in the know or even if its related to the issue. [210375.271752] ------------[ cut here ]------------ This adaptor drop seemed to coincide with me disconnecting an IP camera connected to a different switch, as I had been using the interface, disconnected the camera (3 minutes max) and to come back to see I could no longer ping the adaptors IP. After I restarted the driver all was good again. |
|
Feeling like we're abandoned here.. I don't think the developer has posted in months.. :( |
|
Did you try disabling thermal throttling and enabling Low Power 5G?
In my environment (DS918+), the front port works more stably. |
|
Yes thermal throttling is enabled and here are the current config: Private flags for eth2: The interface does not seem to get overly hot either. I did try enabling low power mode but this causes the speed to drop off below the internal LAN ports. |
How about enabling low power mode with MTU 9000? I tried the same setting in my environment (QNA-UC5G1T / DS918+), and the performance did not drop drastically. So if this configuration is also usable for you, I would like to ask you to check whether there is any difference in stability. |
|
So I will give the MTU 9000 and low power mode a go. Though an MTU of 9000 is not great as not all things on my network like it. As for bandwidth I get the following results with iPerf3 standard MTU, normal power and thermal throttling - 2 - 2.4Gbits/sec |
|
I understand that the MTU 9000 setting is difficult to use in your environment, but did enabling "Low Power Mode" change the stability of your system? I'd like to know the results because knowing this will help me isolate the problem. Just FYI:
kevent 2( If you try to change this setting, use these commands:
|
|
So I have been running the low power mode and 9000 MTU without an interface drop for the past few days. I will continue and report back. I have also doubled the vm.min_free_kbytes from the initial 65536 value and see if the warnings lessen. Many thanks |
|
FYI I still have not had an interface drop since changing to low power mode and using a MTU of 9000. I'll report back again in a few days time. |
|
So still not adaptor disconnects since changing both to low power mode and 9000MTU. |
|
Only solution I have right now is to shutdown and restart my system at 3am every 3 days. :( |
|
So in the past 12 days I've not had any issues at all. I'm just rebooting as its doing an OS update. It would be interesting to know if anyone else trying the low power mode and 9000 MTU also has no issues. |
|
Would like to try it, what are the exact commands to run, and after running them do they "stick" after a reboot? |
Just ssh/putty into the box using a command like Once logged in type to enable low power mode: sudo ethtool --set-priv-flags eth2 "Low Power 5G" on Then change the MTU in GUI. |
Yesterday decided to give it another try. But ALWAYS was using MTU 9000 from the start, and I think experimented with low power also when the issues started. For the moment, 24 hours looks OK. But I definitely rememeber, there were couple of times when link disconnected under heavy load (like I was moving on purpose 100s of GB to desktop) |
FYI It appears the low power mode flag did not survive the OS update though the MTU change did. |
|
So I rebooted my box yesterday to do the OS update and on power up the MTU was still set though the low power option was not and the driver has already died. So I have set it back to low power and we shall see how long it works for. |
|
Thank you for your testing. From this result, I am suspecting this is a power consumption issue. So if you do not prefer to set MTU to 9000 with low power mode, it might be worth trying to use Y-cable or self-powered USB hub with an AC adapter, but I'm not certain. The reason for my assumption is that one of the users had observed voltage drops of USB under heavy load using a USB analyzer and reported that it was resolved by using a Y-cable. I don't know if the voltage drop is due to the load on the host side such as CPU/HDD, or because the dongle consumes more power, or both. But I think it's possible to relate since the QNA-UC5G1T consumes 4.5W of power, which is a lot for a USB device. |
|
Thats really interesting so I have just ordered a USB 3 y cable, with luck should arrive later today and I'll let you know if it helps. Many thanks for the suggestion. |
|
So I got a y splitter cable and connected the power to an external USB power supply, I restarted the NAS, so lost the low power setting so the speed was great, but the the driver has died twice in less then 24 hours. So I have reverted to the low power mode to ensure its solid. |
|
I'm sorry to hear that. Sadly there seems to be no further solution that I can suggest. My DS918+ is working stably, but I don't know what's different between the problematic environment and my environment. |
|
Same here. Tried using connection through USB hub with external power (5V 1A), no luck. Connection dropped within minutes after starting transfers. BTW using 5 HDDs + 1 SSD in my chassis. Maybe some time later will try experiment with another PS - will 2A power will do the magic? Now have to revert to Low power, which gives 2-2,5 GBPS with 1 stream SMB. Knowing this limitation earlier, I would probably go with 2,5G adapter and switch. Here I expected 3,5 G throughput. |
Since the USB device descriptor claims that power consumption is 896mA, a 5V-1A power supply should be enough. But if the device consumes more power due to some kind of glitch, there is the possibility of improving the situation. If it is a problem under high load, it would be helpful if you could try setting the MTU to 16k and see if there is a difference in how often the problem occurs. Increasing MTU will not further improve the throughput, but it will reduce the number of USB packets, which should lower the processing load on the device a bit. It can only be set from CUI: Since this setting consumes more memory, min_free_kbytes should be larger. |
|
I have created a special version that reduces the queue size for USB transfers to suppress the bursty memory allocation. I don't know if this will stabilize the device, but it is expected to reduce the frequency of memory allocation failure errors. I ran CristalDiskMark on my environment, and the errors seem to have decreased somewhat. Try it if you have time. |
|
Many thanks for the updated driver, I will try it out in a few days time after I finish my current tests. So a few weeks back I noticed my SSD read cache (2 x 128GB) had stopped working and was no longer being used. I kept the drives in the unit as it was disabled and thought nothing of it. But a few days ago I investigated a bit more to find one drive was being seen, so after a bit of shutting down the NAS and swapping the drives around I can see only one drive worked. So I removed the dead drive, re-enabled the read cache on 1 x 128gb drive and ever since my ethernet connection has not dropped. Before I could get it go wrong by doing : BTW I have a docker iperf3 server running on the NAS. Since then I have repeated this command from my Mac, using a thunderbolt 10G Nic, ten or more times without any issues so far and I have not reverted to low power mode though I still use a 9000 MTU for both Mac and NAS. I will continue my tests and report back in a few days time and I'm also using the front USB port of the NAS as it was the easiest to get to when I was swapping drives around and as its been working well ever since I have not bothered to change it. FYI my NAS spec is: |
|
A little catchup, so I moved my NAS (I'm moving home soon so I'm trying to tidy things up) so at the same time updated the driver to your test release above. With that said before and since the update I have not had a single drop. I have been really pushing the NAS and as its not on low power mode and running 9000 MTU so I get 3.5Gbits/sec using iPerf3 so I really do not believe I was seeing either a power issue or throughput issue. So it seems removing an unused, but problematic SSD, fixed my issues. Many thanks for all you help. |
|
@sparkplug37, thank you for the status update. It looks like the problem has been resolved. @wdwms and @Globetrotter27, If you still need some help, can you create a separate ticket? I would like to track the reports separately, because if symptoms seem to be the same, the environment and cause may not. @T13nou. have you tried tweaks or the test version of the driver mentioned in this thread? Let me know how it goes now. |
|
I havd same issus
|
|
@iCassius, did you try enabling Low-Power mode? I appreciate it if you create a separate issue. |
|
Hello, Will you do the beta driver who should fix the memory leak for my device? |
|
Hi! |
|
I suspect that in cases where the symptoms do not improve in LowPower mode, memory consumption is not the cause of the problem, but the cause is also the type of device and the cable that connects to that device. However, I am not sure because I have not been able to sort out that information. If anyone with the same symptoms could create an independent issue and fill in the information according to the template, it would help me in my investigation. |
|
In priority to stability, the settings discussed here have been applied by default in recent releases. Since there are so many mixed reports on this issue that it is difficult to track the status individually, we would like to close this issue. |
Description of the problem
Hi, I own a DS918+ with a QNAP QNA-UC5G1T
The adaptater is pretty hot so I've runned the command for thermal throttling.
In any case, same result, the connection drops very often and I need to manually restart the drive from the package interface.
Description of your products
Description of your environment
Output of
dmesgcommandOutput of
lsusbcommandOutput of
ifconfig -acommandThe text was updated successfully, but these errors were encountered: