winbindd - High Memory Usage

[CoffeeBlacker]

The last couple of months a process, winbindd, has been getting larger and larger. I have been checking on it since the QNAP eventually runs out of RAM when this starts to get high. Anyway my understanding was that the winbindd was something to do with samba authentication, which is why i'm posting it here. I've rebooted the system a couple of times to clear that (at one point it got up to 90%+ usage then basically the whole system came to a halt...iscsi connection dropped etc).

Currently at 1206272 K and 50% total usage. Normally my memory usage sits around 25% (SSD caching 256gb seems to use around 15% or so of my available 4GB). We use samba auth for connection from IPADs and other apple products for a series of shared folders for the apple devices. Note i am not seeing high cpu usage until i run out of ram, at that point the system comes to a crawl, our iscsi connections drop...etc.

Note: I have recently made a change to the SSD caching, where i went from Read Only cache, to no SSD cache, then to RW SSD cache. Moving to no ssd cache was horrible, killed my GAVG/latency...Going back to RW cache, has greatly reduced latency for our vm's even compared to RO caching. I wanted to mention this, but also that the original lockup/crash was outside of the scope of any changes made to this. Now i just reboot the qnap when the memory usage gets high (after hours since there are vm's that sit on this...), but i would like to either restart the service when it gets high or preferably resolve the issue.

QNAP TVS-1271u-rp
Pent 3250
4GB
4.2.0 Build 20160311
2 1TB SSDs Raid 1
10 6TB Raid 6
2 256gb SSDs Raid 1 caching

Thanks
CoffeeBlacker

[schumaku]

Hi,

Any specific reason you operate your NAS on an old, outdated, known insecure firmware?

Is this NAS is the default workgroup mode, or is it part of an ADDC?

-Kurt

[CoffeeBlacker]

Sorry for the slow response.

Typically i don't update firmware on devices unless there is a risk or issue with the current firmware, and i don't update to the bleeding edge typically either if i can help it. I'm not sure what you're saying when you say ADDC. It is not a active directory dc if thats what you're asking. It is joined to one of our AD environments and we do use AD permissions for the shares.

More to the question, is there an issue with the Samba service that a firmware update would fix? And this might be asking a bit much, but do you know which firmware update actually fixes it if so.

edit:
current memory usage: 69%
winbindd top mem usage currently at: 1906688 K
next highest: mysqld: 7468 K

[CoffeeBlacker]

What i am reading from this year and last year is that there are greater issues with newer versions of samba on some models, if you upgrade to 4.2.2 x, or even 4.3.x in cases. Instead of dying once a month (or rebooting it once a month), it'd be dead 24/7 depending on what model you're running. Still searching around for answers/information...

viewtopic.php?t=131973
viewtopic.php?t=125875

[CoffeeBlacker]

Additional info...The known issue with the corrent 4.3.3 0174 appears to be very similar to my own issue, but i'm not on 4.3.3 and my issue seems to be currently tied to the process from samba listed above.

https://www.qnap.com/en-us/product_x_do ... 634&II=129
4.3.3 0174
[Known Issues]

The NAS runs out of memory when using thin volumes or LUNs and space reclamation on QTS 4.3.

Note: If your NAS has been online for more than one month, QNAP recommends closely monitoring the memory usage and manually restarting the NAS when necessary. QNAP has identified the root cause and will include a patch for this issue in the next release."

[CoffeeBlacker]

Issues connecting to CIFS mount...similar again but i don't think applies...(forgive me here for listing what may be useless to this thread...i want somewhere where i can go back and look over the info later if need be...and where other people can cross-check if they feel like)
viewtopic.php?f=185&t=132166

[CoffeeBlacker]

Side note about the earlier post on Known Issues, 4.3.3 0174

I am connecting via iscsi to 2 LUNs, one which is block and thick, and another which is file and thick. It specifically says this issue is from using thin volumes or* LUNs and space reclamation on QTS 4.3. I'm not sure again that any of this applies to my situation.

[schumaku]

It does not.