[[6.7.0-rc4] Web Console Stops Working]

another browser didnt help here tooi checked another browser, typed in manually the webterminal address (the adress fiels stays about:blank

i tried from another mashine, i waited till i was at home and tried also, same ...

all other was still working, ssh through guacamole etc ... just the webterminal.

i tried to find a way to restart web interface but afaik theres no way therefore ...

[[6.7.0-rc4] Web Console Stops Working]

same here, no entries in log, just blank screen (about:blank) when trying to open webterminal again.

 

ssh still working etc ...

 

only final solution was a reboot.

[[6.7.0-rc2] New vfio bind fails with multiple devices]

On 1/31/2019 at 7:25 PM, limetech said:

 

This is working by accident because only the first device id specified, 00:14:0, is being bound to vfio-pci and this happens to be your USB controller; and, it's not actually necessary to specify non-Intel GPU's to bind to vfio-pci because Unraid does not include your GPU host driver (at present).

then it was luck  and yes

 

00:14:0 onboard USB controller

07:00:0 PCI-e USB controller

01:00:0 PCI-e Nvidia GPU 1070

04:00:0 PCI-e Nvidia GPU 1030

[[6.7.0-rc2] New vfio bind fails with multiple devices]

just to be sure, my BIND line would then look like this ?

 

BIND="00:14:0 07:00:0 01:00.0 04:00.0"

instead

BIND=00:14:0 07:00:0 01:00.0 04:00.0 <<-- wich i currently used and seemed working, both USB and GPU are passed through and working

[Unraid OS version 6.7.0-rc1 available]

9 hours ago, limetech said:

Correct.  I would be interested in your results doing exactly that and verify that everything works correctly.  re: the Signal processor: It is my understanding that for proper operation all the devices of an IOMMU group must be isolated (bound to vfio-pci) or none of them.  Maybe you can move that USB controller to another slot in order to put into its own IOMMU group?

ok, i did it like described and all working fine here, both vm´s running smooth with nvidia and their usb controllers

 

only thing not working seems tobe the openvpn server plugin but thats surely not related to this.

 

about splitting, its one of the onboard ones ... so no chance to move into another slot  (i have 2, 1st for unraid usb key, 2nd for workstation vm) i cannot isolate it but i see all running here ...

 

i attached my system log in case of interest ...

alsserver-syslog-20190122-1831.zip

[Unraid OS version 6.7.0-rc1 available]

im also guessing what todo

 

updated and everything seems to work so far, passthrough nvidia and USB onboard without changing anything yet.

 

currently i have this in my flash config

 

kernel /bzimage
append pcie_acs_override=downstream isolcpus=2,3,4,5,8,9,10,11 vfio-pci.ids=8086:a2af vfio-pci.ids=1106:3483 initrd=/bzroot

 

when i read this correct i remove the vfio part here and create a file called vfio-pci in /config

 

then i would add those 4 entries to vfio my 2 usb controllers and 2 nvidia cards ?

 

BIND=00:14:0 07:00:0 01:00.0 04:00.0

 

so these should be free for VM´s, just a little worried about IOMMU group 5 cause the Signal processing ... is included ...

 

IOMMU group 5:[8086:a2af] 00:14.0 USB controller: Intel Corporation 200 Series/Z370 Chipset Family USB 3.0 xHCI Controller

[8086:a2b1] 00:14.2 Signal processing controller: Intel Corporation 200 Series PCH Thermal Subsystem

 

IOMMU group 20:[1106:3483] 07:00.0 USB controller: VIA Technologies, Inc. VL805 USB 3.0 Host Controller (rev 01)

 

IOMMU group 15:[10de:1b81] 01:00.0 VGA compatible controller: NVIDIA Corporation GP104 [GeForce GTX 1070] (rev a1)

[10de:10f0] 01:00.1 Audio device: NVIDIA Corporation GP104 High Definition Audio Controller (rev a1)

 

IOMMU group 18:[10de:1d01] 04:00.0 VGA compatible controller: NVIDIA Corporation GP108 [GeForce GT 1030] (rev a1)

[10de:0fb8] 04:00.1 Audio device: NVIDIA Corporation GP108 High Definition Audio Controller (rev a1)

 

[disk spinup since ~ 6.6.x]

after some tests with disabling cache_dir plugin i can confirm its back to normal again.

 

What i didnt test yet is a reboot without cache_dir plugin (uninstall, reboot, check ...), only disabled now ...

 

as there where no changes still a myst what made this happen

[disk spinup since ~ 6.6.x]

after 2 days usage with plugin cache_dir and pressure 0 im there where i can say its normal like before ...

 

drives only come up when they should, no more spinups by laptop booting, no more spinups by access media files from cache ... like it was before latest updates ...

 

i hope that this "workaround" wont break other things, but now im good atm.

 

i only have one spinup @ nite ~ 2am without my personal use, but i guess it has something todo with plex, wich was also before latest updates ... so no change here but also doesnt matter.

 

just as note if that may points in the direction what could have changed and causing disk spinups.

[disk spinup since ~ 6.6.x]

i increeased now the following in the plugin cache_dir, cache pressure 0, ulimit (memory) 0 ...

 

wich also improves the behavior of silent disk usage.

 

may anything to provide wich could be helpful ? could it be the smb 3.1.1 update ?

i mean all is running smooth here, just disk behavior changed ...

 

and maybe (just a feeling), when starting media´s in clients, the startup takes longer .... even now when the disks

does not all spinup (due cache_dir plugin), like it would take longer to server the data ... from the cache,

specially the 1st playback, sample user case, i start a show, it takes about 5 - 10 seconds to startup wich has been faster before when coming from cache ...

when i know use /mnt/cache instead the media starts instantly ... same file, same location.

i just try to figure where to look for cause, in my eyes i cant find anything what could be the reason, no log or any watch plugin (active streams, file activity, open files) shows a reason why a disk is now starting up ...

 

so in combination with "feeled" access times i think about the access behavior to /mnt/user ... specially on files wich are in physically in /mnt/cache/...

 

so if theres any usercase u need logs, i ll try to provide them, i cant find anything and i dont know how to describe it better

[disk spinup since ~ 6.6.x]

7 hours ago, trurl said:

Does that user share include all disks?

yes it does

[disk spinup since ~ 6.6.x]

ok, some othe tests i made

 

i record tv shows with plex, all files are physically in cache

 

when i now open Avisynth to cut my commercials

 

by just starting webui of avisynth and accessing a folder ALL disks spinup without touching any file,

tis happens when using /mnt/user/Media as start folder, and im pretty sure Avisynth also doesnt read all files ...

also in this folder there definately no files from ALL disks, usually files are in /cache or /disk4, im ONLY working

with the files on /cache, but ALL disks get active ...

 

when i use /mnt/cache/Media im good of course, so i still believe (cause i dunno better) that unraid has some issue

when accessing /mnt/user/... to recognize there are cache files, also here im pretty sure that wasnt like this before.

 

and again, NO activity in open files, in active streams or any other log ...

 

so even IF in this case Avysynth would look for the files in the directory im currently in, it shouldnt fire up ALL disks, at least in my understanding of cache and disks ...

 

media access by plex server btw got almost 0 activities now by using the plugin cache_dir with fixed value und pressure 1, so here i have a workaround wich helps alot, but other access is still ... lets say depends on server mood  sometimes ok, sometimes just starts all up with no (known) reason.

[disk spinup since ~ 6.6.x]

ok, made a safe mode test now

 

started with GUI safe mode, started the array, stopped all dockers, no plugins where running.

spinned down the drives so only cache (NVME) and UA (NVME) was active.

 

then started my laptop wich has the shares mounted.

 

2 of 3 times by only opening the explorer on the smb client the disks starts to spin up ...

when it kept calm, entering the network share 1 of 2 will spinup the drives

 

accessing media wich is in /mnt/user/Media/test ... and is physically on the /mnt/cache/Media/test ...

will sometimes spinup the drives.

 

so i NEVER executed or touched (opened) any file wich is really on one of my disks (disk 1 - 4).

 

i guess that the smb accesses (even when "sleeping") from clients (VM´s, Dockers, Shield TV, Kodi, ...)

are resulting in spinups, even there are no files executed or touched.

and even when a file is executed from cache inside the same share where disk files are ..

 

sample

/Media/Show/Season1/Episode1.mkv   <-- is physically on /server/disk1/Media/Season1/Episode1.mkv

/Media/Show/Season1/Episode8.mkv   <-- is physically on /server/cache/Media/Season1/Episode8.mkv

 

when i now start Episode8.mkv on an plex client or kodi client, disk1 will most likely spinup ...

so unraid (smb) doesnt really know where the file is in 1st case and look in all disks ...

wich im pretty sure wasnt like that ... and shouldnt be like that.

 

i think when i browse the shares in the webui from unraid i never get any spinups ...

only when i really hit the file and a download or whatever starts then the disk will get spinup, wich is correct.

thats just my thoughts.

 

as i never saw any entries in syslog, file activities plugin, open files plugin wich could explain what is happening ...

i made a diagnostic from the last test today morning, but i cant see anything, hopefully u can tell what could be the reason ...

 

alsserver-diagnostics-20181101-0756.zip

 

[disk spinup since ~ 6.6.x]

ok, so on boot, i know that option.

 

may will look on this tomorrow when i get some more time, but actually some other behaviors

 

from my win10 VM

 

accessing the same video files directly from win 10 with open in VLC

 

/server/cache/Media/.... -> no spinups

/server/Media/... -> disks will spinup after 2 or 3 files ...

 

ALL tested files are physically in /cache, just accessing them differently ...

 

now, im considering to split my Media Points so i always have 2 sub dirs, /mnt/cache/Media... AND /mnt/Media ....

or even /mnt/cache... /mnt/disk1/... and so on, would need some scraping again

then at least media access should be solved (wich is a workaround and no solution in my oppinion)

 

same behavior also with other "apps", sample, yesterday recorded Walking Dead, opening AviSynth for cutting, disk spin up ... also in cache, the complete season 9 yet  but i ll take a look tomorrow in safe mode.

 

i just dont like to reboot cause boot time from my 10G NIC is insane long (2,30 min)

[disk spinup since ~ 6.6.x]

and some more tests, when i playback files in plex or kodi for example wich are definately on the cache drive,

also all disks spinup before the playback starts ...

 

im pretty sure this behavior wasnt before, before when i played "actual current media" it started instantly (from cache), when i started playback after the mover turned in (was once a week) i could here playback started after the disk spinup like it should be ...

 

now i always have disk spinups, nevermind where the media file is located ...

 

just tested 10 files and always disks spinned up while all files where located at /cache ...

 

i ll test some more, sadly i cant see anything accessing those disks in any logs ...

[disk spinup since ~ 6.6.x]

well, when u mean safe mode from boot on, i never tested this yet.

 

or do you mean by disabling all plugins, dockers and vm´s ?

[Unraid OS version 6.6.0-rc1 available]

ok, thanks for the info, then i stick with pipework for now.

54 minutes ago, ken-ji said:

Unless something changes with the Linux kernel or the docker engine, this isn't going to happen. Maybe if we had the ipvlan drivers... but this has its own caveats and is not yet in prime time AFAIK.

 

[Unraid OS version 6.6.0-rc1 available]

upgrade worked here fine it seems, one question i have about igpu

 

i915.alpha_support=1  <<-- still needed to get hardware modprobe i915 working on coffee lake cpu´s ?

or can this be "updaated" to 965 (no idea if possible)

 

and is it now possible to give dockers their own IP´s without isolating them from the host network ?

 

thanks ahead