Re: Using "Enable fast CIFS writes" may corrupt OS on Sparc.

grunnsat · ‎2014-10-19

I'm using a ReadyNas NV+ with 1GB of memory. Ever since Radiator 4.1.12 and 4.1.13 I had the problem that after a couple of weeks Radiator got corrupted. That was visible in Raidar and Frontview, because both did not show the Radiator version number any more. Shutting down the ReadyNas from Frontview wasn't possible either, I had to use the on/off button. The only way to fix the problem was to re-install Radiator.

I also noticed that the disks did not spin down any more, every few seconds I saw the disk activity LED flash.

It occurred to me that something might overwhelm the processing capacity of the ReadyNas, and that could corrupt Radiator. After checking all the Frontview settings I decided to disable "Enable fast CIFS writes", and ever since the disks do spin down, and I haven't seen the corruption any more.

It hasn't been that long ago since I made this change, but I thought I let you know anyway.

StephenB · ‎2014-10-19

I'm running 4.1.14 T6 on an NV+ v1 and a duo v1. Enable fast CIFS writes is enabled, as is the disk cache. Journaling is disabled (and I have a UPS). Both have stock memory.

Disk Spindown works, and I have had no trouble with OS corruption.

The obvious difference I see is that you are using 1 GB ram and I am not.

grunnsat · ‎2014-10-19

That is why I mentioned the 1GB memory. I'm sure you know there is a problem with the Readynas Sparc systems with 1GB memory, and it seems to me that the fast CIFS writes setting is causing the problems. I will watch what the ReadyNas does, but so far it does look like this is the problem. Perhaps someone can check the code of "Enable fast CIFS writes" and see if he can find anything that can explain the corruption.

grunnsat · ‎2014-11-13

I have been running firmware version 4.1.14 for about four weeks now, and I disabled the "Enable fast CIFS writes" setting, and the firmware corruption problem has not appeared again.

It was reported that the "Enable fast CIFS writes" actually doesn't do anything any more, is this true? Anyway, I would recommend anyone using a Sparc ReadyNas with 1 GB memory to disable this setting!

grunnsat · ‎2015-03-06

My NV+ has been up for months now, and the firmware corruption problem has gone. As far as I'm concerned, that proofs that using the "Enable fast CIFS writes" in combination with 1GB memory will corrupt the OS. So you should not use this setting if you have 1GB installed in a Sparc based Readynas.

I would love to see a comment on these findings by one of the moderators if possible.

StephenB · ‎2015-03-06

I have no inside knowledge here (as I don't work for Netgear).

Though I think what you've proved is that there's an issue with your memory pure and simple. Visible symptoms may have disappeared when you stopped queuing data in memory by turning off fast writes. But I don't think that actually fixes the underlying problem. It could well show up in other more subtle forms that are harder to detect.

Also, I don't think you've proved anything (either way) about 1 GB ram generally. Some systems might work fine. (We do know from many posting that the sparc platforms are very finicky about their memory, which is why Netgear eventually stopped supporting upgrades).

If you still have the stock memory, I'd suggest reinserting it (and re-enabling fast writes). RAM is a key component of the server - if there's evidence that it isn't working reliably I'd be very uncomfortable about trusting my data to the device.

grunnsat · ‎2015-03-06

I've been running this NV+ with 1GB almost from the beginning, and never had any problems until 4.1.12 if I remember correctly. From that moment on I had these problems, and others had problems as well. I ran the memory tests several times, never saw a problem. So I doubt very much that this is a hardware related problem. In my view it is more likely that 1GB of RAM is simply to much for the cache driver.

I was also told on this forum that the "Enable fast CIFS writes" doesn't do anything any more, which is odd since disabling it solved my problem.

StephenB · ‎2015-03-06

Well, I think you have hard evidence that "Enable fast CIFS writes" does something.

I get your thinking, but the conclusion you are reaching on the cache is basically speculative. You could be right, but maybe not.

grunnsat · ‎2015-03-06

Another possibility is that the functionality of "Enable fast CIFS writes" has somehow been incorporated in the standard software, and that "Enable fast CIFS writes" now does something undesirable, at least with the 1GB systems.

You're right of course, I'm speculating. But I'm trying to find a logical explanation. My brother has a Duo with 1GB, and his system was so corrupted that only with all kind of tricks I was able to save the data and get it running again. It is a serious problem ........

StephenB · ‎2015-03-06

Well, if it no longer is supposed to do anything they should have removed the option from the UI.

I agree it's serious btw. The thing I'd worry about is that disabling fast CIFS writes might not be enough.

grunnsat · ‎2015-03-07

I agree, if it is true that the "Enable fast CIFS writes" setting is not suppose to do anything any more, then the setting should have been removed.

After disabling the setting, my NV+ has been up for four months without any problem. The disks spin down when there is no activity (that was a problem too), and yesterday I did a volume check and no problems found. I'm quite confident that there are no more problems now.