Memory issues for RN104 ?

[espie]

I've recently upgraded my firmware to 6.9.0, then 6.9.1, and I've added disks up to 4 x 3TB (default raid mode, one volume, so one disk for redundancy)

I'm currently experienced some memory issues with the RN104. It's a fairly simple setup, mostly used for storing video files. I have minildna and nfs running, and that's it (okay, also the default httpd admin setup)

Since the logs don't give that much information, I've logged in thru ssh so that I can look at top and the systemd journal.

After a while, almost always after writing stuff thru NFS, I see the NAS swapping... eventually killing minidlna (I know how to restart it, thank god).

And eventually, the kernel crashes... the one line summary on the display isn't enough, often it's in putname+64.

So I've taken to rebooting once the  swap is entered, which sucks.

I'm somewhat annoyed, because I'm well beneath the maximum advertized capacity of the RN104.

Since I added the 3rd disk at about the same time I moved to 6.9.*, I'm not quite sure whether the firmware update or the new disk is the culprit.

I've got several questions:
- is there a way to reduce the memory footprint of what's running, even if that means getting a bit slower. In particular, I see a lot of nfsd and kworkers. Where is this configured ?
- I could get rid of the httpd admin. I know how to restart it thru ssh + systemctl, probably.
- has my system gone bad ? Is there a way to reinstall things while keeping my data ?  I don't really have the money to buy a second system and do a full backup.

- if the system is indeed too small for what I'm trying, as far as I know, there's no way to add memory to a RN104.
Is it possible to buy a newer netgear nas, with 2GB, and just pop the disk in without losing data ?
I could possibly buy a newer empty nas, but new disks would put me waaay over budget.

[espie]

Well, I found where the number of nfsd threads was hidden... Assuming this consumes quite a few buffers, I've throttled it down from 8 to 4, which ought to be quite enough for my usage.

Let's cross fingers and hope this helps.

I have no ideas which of the various btrfs/raid tasks are memory hogs on this machine.

[espie]

Three days later, my NAS hasn't gone into swap!!!

This ought to be exposed in the admin interface on the web.

NAS with 512MB are now short-pressed for storage, but those nfsd threads do consume a lot.

If your NAS is forced into swap, then the performance will suffer a lot, and it will crash eventually (or various services will fail).


The log/admin interface is not enough.   Out of memory kills should be visible. You should be able to query the status of running services, or even restart them (systemctl query  service).  You should be able to see the same memory information basic top shows (at least memory/swap).   minildna is 100% flaky once you hit swap (devices like the xbox will error out while playing videos)

And you should be able to change the number of nfsd threads.  8 is complete overkill for a "personal nas", especially one with a full disk array, and only 512MB.

Took me a few hours to sleuth this out, because I'm not a linux person (all my other machines run OpenBSD).
The idea of the NAS is that you shouldn't have to ssh in as root to fix this.   Even the dumbed down web interface should show this.

Please consider it for a further update...

Another thing that would actually be cool would be to offer an option to install documentation on the NAS thru the admin.  Yep, it takes a bit of space on the disk. But having "man" not return anything for any command sucks.  I know you can get those on the web, but this still sucks.

[StephenB]

---

@espie wrote:

NAS with 512MB are now short-pressed for storage, but those nfsd threads do consume a lot.

---

Yes, that is a problem on the RN100 series.  Thanks for sharing your solution here, I suspect others are also using NFS.

---

espie wrote:  And you should be able to change the number of nfsd threads.

---

That is something it supports now, on system->settings->services->NFS. 

---

@espie wrote:

The log/admin interface is not enough.   Out of memory kills should be visible. You should be able to query the status of running services, or even restart them (systemctl query  service).  You should be able to see the same memory information basic top shows (at least memory/swap).   minildna is 100% flaky once you hit swap (devices like the xbox will error out while playing videos)

---

I agree that it would be useful to see more information - particularly CPU and memory loading.  Perhaps post an idea in the idea exchange on this.

Maybe also suggest lowering the default setting for nfsd threads.

[espie]

Oops, my mistake, did not find the nfsd settings in the interface 🙂

All my other remarks still apply.

[StephenB]

---

@espie wrote:

All my other remarks still apply.

---

I agree.  Netgear development is more likely to consider them if you repost them to the idea exchange.