I am not a programmer. I was wondering, how does one change the lowmem_reserve_ratio? How do I access that? Not sure where to find those parameter lines. Running Windows 7 Home Premium.
Thanks.
Thanks _ agerstla _, the only thing that has made my WD My Cloud 4TB stay up for longer than 8 hrs was setting the lowmem_reserve_ratio.
echo '16 16' > /proc/sys/vm/lowmem_reserve_ratio
It’s been a very frustrating time trying to troubleshoot this network disconnect problem. I tried all the other fixes: setting up a cronjob, disabling sleep, etc., but nothing changed. And sure, it’s only been 2 days now, but that’s about six times longer than it’s been up since I bought it at the end of December.
Thanks!
@starhopper: This requires you to log in via ssh and run the ‘cat’ command that was given. If you don’t know how to do that, you may not want to fiddle with it. Doing something wrong can really brick the machine.
@Aerodoq and others: glad it has helped. I am still seeing allocation failures in the logs on my machine. They all seem to be rooted in the network stack (tcp_xxx), which explains the connectivity and transfer speed issues. Went up to ‘32 32’ but that doesn’t seem to help much. In the end, there is probably some crashing one has to live with. The problem is also the Linux kernel itself. Apparently it doesn’t work optimally on embedded machines such as this one that do not have a memory management unit (MMU) built into the hardware/chipset itself (but want to run memory-intensive jobs). That really needs to be improved in the kernel, and with Android being based on similar setups, I assume there is active development (the latest firmware runs a 3.2 kernel and there are newer versions out there, but I am not that adventurous).
I am going to return my unit to Newegg immediately. This “MyCloud” is a great idea, but is still too buggy. If we accept poor products, we will continue to receive poor products. I will not simply “live” with this issue and hope it’s fixed someday.
If enough of us refused to accept shoddy workmanship and returned these, perhaps Western Digital would finally feel the need to fix the issue.
You guys have fun!
agerstia,
Thanks for your help. My disconnect is not happening at times of high transfer - rather, overnight when there is no activity. I have had to unplug and restart mycloud every morning since I’ve had it operating (4 days)
I am returning it to Newegg immediately as I refuse to accept less than advertised perfomance. I am not a computer programmer - and this thing needs to be “plug and play”, not “plug and plug and plug” :smileyvery-happy:
Perhaps, if we all returned these units and refused to accept this, WD might actually do something about it.
Cheers
Mine did this initially, but it has been perfect ever since the last firmware upgrade. Even the DLNA access time seems to have improved recently.
Hi guys, this issue was resolved on the latest firmware (3.03.01), be sure to do the update if you are experiencing any connection problems.
Firmware Release 3.03.01 (1/7/2014)
- Resolved issue of drives randomly losing connection to the network.
Captain_Panaka wrote:> Other NAS products have 512M and manufacturing cost would increase by 4$ max.
Unbeleivable!
Other NAS products have 128M and work fine. (Pogoplug, for example.)
4-bay QNAP TS-4xx NASes have 256M of RAM and work great.
4-bay Drobo NASes have 256M of RAM and work great.
4-bay Synology DS-411 have 256M of RAM and work great.
Unbelievable!
I’m curious – how do you know with such certainty that desigining in another 256M would only cost $4?
You have to think about manufacturing costs, supply procurement, resources, etc for hundreds of thousands of units. This in turn, increases the price point. Would you be willing to shell out more money than what you paid for an extra 256M? Say some where along the line of an extra 50-100$ (just assuming)
I have the latest firmware update - did nothing to resolve the disconnect issue.
Guys - found out how to get a response from WD Tech support! I posted a very negative review (gave it one star) on Amazon.com - and within 12 hours received this reply…
"Please allow us to apologize for the continued inconvenience you are experiencing with the My Cloud and loss of connectivity. To rule out any possible issues with a DHCP lease between your router and the My Cloud, please visit the link below which will provide you with instructions on how to set a Static IP for the drive, along with configuring your router for remote access to the drive.
“How to enable port forwarding on a network router for use with a WD My Cloud, My Book Live, or My Book Live Duo”: http://wdc.custhelp.com/app/answers/detail/a_id/8526
Our support team would like the opportunity to follow-up with you on this, should the issue persist. We are able to collect logs from the drive’s OS as well as perform a network scan to look for anything that might be triggering the symptoms you have stated, and if necessary escalate this for further review. Please contact us via email or phone at the below information and mention this Amazon review when you speak with our My Cloud support.
(855) 556-5117
http://support.wdc.com/contact/contact.asp?lang=en&ct=sentinel
So, I will take their advice and try this again.
@starhopper: The disconnects happen at night because that is when the media scanner and twonky server automatic scans kick in. Those are memory hogs (it is not transfer activity so much as memory load that triggers the page faults). This also explains why some people can get better stability when turning off the various media, DLNA and itunes server functionalities in the box. (BTW, you may want to try that). But I can see your point, a product should work as advertised out of the box and not require fiddling on this level. For me, just call it professional curiosity of poking around what is going on.
@jubei04: The latest firmware definitely does *not* fix the disconnects. As I wrote in my original post, I am on the latest firmware. When I bought my box, the first thing it did was update itself. The second thing it did was disconnect from the network once I started to transfer some files into it.
There a couple of posts on here re overnight scanswhen lots of new files causing it to vanish?
I put about 500Mb of data on immediately after buying it, several days worth of near continuosu file transfers and syncrionisations asmy previous backup was full. It always stayed live during the transfers, but then vanished from the network the next night even with the blue light still on.
Now a few weeks from doing this later it is much faster to access the files etc and hasn’t disconnected since the firmware update? I wonder if it just takes a very long time to catalogue everything.
It may take a long time for it to catalog everything, but it should never vanish.
If it’s going away for any period of time, something’s wrong somewhere.
Well, I turned off “Media Streaming” and “Itunes Streaming” and the mycloud has not disconnected itself in 24 hours. I will hold my breath for a few more days and see if it continues to work.
Since I don;t need those services, it’s not a big deal TO ME. However, it is a poor design if one has to disable major features of a device to get it to stay connected to the network.
Shame on WD.
Another update: after making the change to lowmem_reserve_ratio, my box has been up for 4 1/2 days where its previous record was 8 hours.
However, I have also noticed a near constant load average of just over 3. Anyone else notice a consistently high load average? Watching top doesn’t show any processes above 10% CPU, but twonky and wdphotodbmerger seem to be running anywhere between 3-7% consistently.
A>
Just a quick update, after some experimenting with different values, the best configuration I found on my box is:
echo “64 32” > /proc/sys/vm/lowmem_reserve_ratio
I still get faults, but the least with that setup it seems (but that will likely depend on a lot of factors).
I am also seeing that ‘top’ reports a constant load of around 3. This seems normal. But I think there is a bug in the compuation of reported load values. Given what I see running, that doesn’t look like a load of 3.
@ Captain_Panaka: Memory management in Linux (well, in general) is complex. See the link I posted earlier for one of the better explanations I found (better than I could explain it). Another explanation is here:
http://utcc.utoronto.ca/~cks/space/blog/linux/DecodingPageAllocFailures
http://utcc.utoronto.ca/~cks/space/blog/linux/WhyPageAllocFailure
The latter also talks about some root for the problems we are seeing (some would call it a bug) in the way the Linux kernel is written.
agerstla wrote:
I am also seeing that ‘top’ reports a constant load of around 3. This seems normal. But I think there is a bug in the compuation of reported load values. Given what I see running, that doesn’t look like a load of 3.
agarsta: Yeah, that’s “normal.”
The reason the load is at 3+ is because there are a few kernel processes running that cannot use hardware interrupts. They’re watching the reset button, controlling the parallel core (specific to ARM architecture) etc. They’re constantly polling things, then go to sleep. Since they’re in “uninterruptible sleep” in software, so they’re still queued by the CPU.
That’s why you’ll see 99% idle but still have load of 3.
Or something like that. :)
@agerstla What differences did you see between the ‘16 16’ and the ‘64 32’? Mine has been stable with the ‘16 16’, but I’m willing to entertain the idea of changing it again if there are some tangible benefits. 
A>
