• 1
  • 2
  • 3
  • 4(current)
  • 5
  • 6
  • 11
NFS issues continue with Gotham beta 2
#46
Sure. I'm just attempting to clarify how we got to where we are now. Not slate anyone or start a flame-war.

Smile
Image
Reply
#47
I just checked my ReadyNAS and I have it set to 4 NFS Threads. Now I run XBMC on 4/5 Devices regularly. 1x HTPC, 1x Desktop, 1x Laptop and 1x Nexus 7 (2012). However Usually, this is on a single device at any one given time. Load is also spread across the 2x NAS devices I have (1x ReadyNAS Pro and the primary server which is a modded ReadyNAS Repertoire with an Ultra motherboard). So no more than two devices are ever likely to be using NFS at any given time.

The Repertoire/Ultra does have Plex running in the background on it too, but this is primarily for remote access. The Pro has Subsonic running on it (also for remote access) and also my downloader apps.
Image
Reply
#48
(2014-03-28, 22:52)gibxxi Wrote: "A Few Users"? Define A Few? I've not been following the Beta thread all that closely, but I can't help thinking downgrading NFS for the sake of "A Few Users" is a bit of a knee-jerk reaction. Is it even confirmed that those users issues were directly caused by the newer libNFS library?

One could easily make the same argument in reverse. That is, there's at this point only 2 or 3 users complaining about the artwork issue.

No offence but you all came to this quite late and hindsight is 20/20.

A problem was identified - a very serious regression - of total library dumping, and it was isolated to the libnfs bump (which wasn;'t so hard as it happened Frodo 12.2 to Frodo 12.3 and there weren't many changes. A few users (me, at least 2 others) chimed in in the main thread and on github. We tried wiresharking, nfs logs, etc., and a few other things - Memphiz did heaps. Unfortunately it didn't work, so a decision was based to go back to the much more thoroughly used and tested libnfs 1.3 after I popped it in and it stopped the nuked library issue. I did not notice the artwork issues, presumably because of caching etc and because I was very focussed on that one main issue. Whichever way you cut it artwork issues are less important than library wiping ones. Date for both is understandably not much as there are lots of setups and overall nfs on Windows, and people running betas with it, is going to be a low number of people.

Anyway, none of it f*cking matter really as - here we are.

The best thing for everyone would be be to solve either issue - either the outright library dumping with newer libnfs on windows (tested against two servers on setups working with the older libnfs for *years*), and I guarantee you the ones that reported it so far won't be the only ones that hit this - OR solve the artwork issue with the older libnfs. Either way is a good move forward - the first would be preferably of course but we'd need someone who can (quickly!) compile and work on libnfs on windows (I am happy to do any testing and could probably get it compiling but I know zero about nfs internals really).

Or - as is probably the most likely right now - they may well simply remove nfs for Windows. Which is also definitely not ideal.


*** One thing to note - I confirmed last night that beta1 with the libnfs1.3 DOES seem to have the artwork issue - implying the artwork issue started *before* beta 1. @garretn - do you want to try confirming this maybe?
Addons I wrote &/or maintain:
OzWeather (Australian BOM weather) | Check Previous Episode | Playback Resumer | Unpause Jumpback | XSqueezeDisplay | (Legacy - XSqueeze & XZen)
Sorry, no help w/out a *full debug log*.
Reply
#49
Isn't the artwork issue a known fault of libnfs1.3, fixed in libnfs1.6?

If so, surely the only way forward is to figure out why 1.6 is causing issues with the nfs shares going AWOL (or whatever is the issue).
Always read the XBMC online-manual, FAQ and search the forum before posting.
Do not e-mail XBMC-Team members directly asking for support. Read/follow the forum rules.
For troubleshooting and bug reporting please make sure you read this first.


Image
Reply
#50
Well the artwork never went awol with Frodo and libnfs 1.3 but ultimately yes. No question - working out why nfs shares/stat fails with latest/current libnfs would definitely be the best thing. Problem is finding someone who can do it - I've been a coder but not that sort/level. All server side java enterprise stuff and php/python web stuff. I jsut don't have the C/C++ to get it done.

I am happy to work very closely with anyone that can do it - and jmarshall we're on roughly the same timezone Wink
Addons I wrote &/or maintain:
OzWeather (Australian BOM weather) | Check Previous Episode | Playback Resumer | Unpause Jumpback | XSqueezeDisplay | (Legacy - XSqueeze & XZen)
Sorry, no help w/out a *full debug log*.
Reply
#51
I don't have (nor desire to have) a windows machine, so you won't get help from me I'm afraid.

I would think that the best plan of attack would be to figure out libnfs building for yourself, then add a shitload of printf() statements in the stat() routines.

Cheers,
Jonathan
Always read the XBMC online-manual, FAQ and search the forum before posting.
Do not e-mail XBMC-Team members directly asking for support. Read/follow the forum rules.
For troubleshooting and bug reporting please make sure you read this first.


Image
Reply
#52
@bossanova808, calm down buddy. Nobody was trying or intending to belittle the library issue, and I *never* said the artwork issue was in Beta1 (Because Beta1 uses 1.6).

Like you said, it doesn't matter. The reason I'm not interested in fixing 1.3 is because that's fixing a fix, it's far more ideal to fix the original problem instead and get all platforms on the same libnfs version again.

Now, lets move on. I'm still working on this, and I'm starting to think that the duplicates problem mentioned may actually be related but at this point is still a suspicion since I'm having a little trouble getting libnfs to compile (so I can add some debugging et ceteras). I suspect the attributes used to identify and compare files is a potential culprit, resulting in duplicates -- and also what would appear to be a missing file to the cleaning functions since those don't actually use the file location alone, but also hashes.
Reply
#53
(2014-03-29, 02:36)garretn Wrote: @bossanova808, calm down buddy. Nobody was trying or intending to belittle the library issue, and I *never* said the artwork issue was in Beta1 (Because Beta1 uses 1.6).

Like you said, it doesn't matter. The reason I'm not interested in fixing 1.3 is because that's fixing a fix, it's far more ideal to fix the original problem instead and get all platforms on the same libnfs version again.

Now, lets move on. I'm still working on this, and I'm starting to think that the duplicates problem mentioned may actually be related but at this point is still a suspicion since I'm having a little trouble getting libnfs to compile (so I can add some debugging et ceteras). I suspect the attributes used to identify and compare files is a potential culprit, resulting in duplicates -- and also what would appear to be a missing file to the cleaning functions since those don't actually use the file location alone, but also hashes.

Agreed. FWIW Bossonova808, i've not had this issue, as I've already stated. However that doesn't mean it may not rear it's head in the future. So while the library artwork issue is less than desirable, having a working XBMC library (with or without NFS support) is certainly more important. However, While those of us that have noticed the issue with NFS support and library artwork this early on may be small in number, that certainly will increase tenfold once Gotham goes gold. The forums could probably do without scores of disgruntled Windows users moaning as to why their artwork isn't showing with NFS.

Me? I'll just continue to use SMB or find a workaround if that's what it takes. But it may be a wise choice if garretn can't find a solution, or indeed the two of you can't find a workable solution, to flat-out remove NFS on Windows for the final version. It may save a lot of headaches down the road. Just an idea.

I'm prepared to test any fixes going forwards, but the value of any testing I may do will be limited since I wasn't suffering from this issue to begin with. But thanks to those who are (and have been) looking at it, regardless.

Smile
Image
Reply
#54
(2014-03-29, 02:36)garretn Wrote: I'm having a little trouble getting libnfs to compile

This is the code you are looking for https://github.com/Memphiz/libnfs/commits/xbmcgotham (thats where our 1.6.3 package of libnfs is made from)
AppleTV4/iPhone/iPod/iPad: HowTo find debug logs and everything else which the devs like so much: click here
HowTo setup NFS for Kodi: NFS (wiki)
HowTo configure avahi (zeroconf): Avahi_Zeroconf (wiki)
READ THE IOS FAQ!: iOS FAQ (wiki)
Reply
#55
bossanova808, do you think you could set up sqlite as your database (if not already), and provide before and after copies of the MyVideos## database from when the files are there, and directly after cleaning removes them?
Reply
#56
I could do that I guess. I use mysql typically but I could disable that and maybe scan a subset of the library to keep it all small, quick and manageable.

To what end, though? Have a theory?


I take it you've looked at the logs etc. on github (xbmc logs and wireshark?) - basically it seems the stats fail and then xbmc does a mass delete query.
Addons I wrote &/or maintain:
OzWeather (Australian BOM weather) | Check Previous Episode | Playback Resumer | Unpause Jumpback | XSqueezeDisplay | (Legacy - XSqueeze & XZen)
Sorry, no help w/out a *full debug log*.
Reply
#57
Add printouts in xbmc/filesystem/NFSFile.cpp - NFSDirectory.cpp Method "Stat" and verify that its really the cause!
AppleTV4/iPhone/iPod/iPad: HowTo find debug logs and everything else which the devs like so much: click here
HowTo setup NFS for Kodi: NFS (wiki)
HowTo configure avahi (zeroconf): Avahi_Zeroconf (wiki)
READ THE IOS FAQ!: iOS FAQ (wiki)
Reply
#58
Something like that. I think the stats failing (well, if thats whats happening) might not be exactly whats going on, just "close" to the issue.

Rather, I was suspecting it had something to do with the hashes/fasthashes and the earlier mention of the duplicate issue. My duplicate issue I wasn't taking too seriously since the test that caused it wasn't a particularly good test, but then others mentioned also having that problem with 1.6. If the hashes are changing, then the old files would appear to be missing -- so it'd clean them, and the same issue could cause duplicates. That's the theory, at least.

That code linked for building libnfs worked perfect btw, I just haven't had time to do anything with it yet unfortunately.
Reply
#59
@garretn @Memphiz

I will try re-surrect my xbmc build environment but happy to try and libnfs.dll and/or xbmc with printfs added if you beat me to it.

I'm very busy at work right now but hopefully I can find a chance to stay up late and get hacking.
Addons I wrote &/or maintain:
OzWeather (Australian BOM weather) | Check Previous Episode | Playback Resumer | Unpause Jumpback | XSqueezeDisplay | (Legacy - XSqueeze & XZen)
Sorry, no help w/out a *full debug log*.
Reply
#60
I haven't read every post but is it in the plans for Gotham to fix nfs or is there a workaround? It's not critical as both my machines run OpenElec but my PC has beta 3 on it and I can't access any media. I had thoughts of using it as my scraper but the broken nfs prevents that. I could erase the database and go back to SMB but I've so many positives with nfs.
Server: Synology Diskstation 1511+ with 8x WD Red NAS 3TB drives, DSM 5.2
Main HTPC: Home Built i3, 8GB RAM, Corsair 128GB SSD, nVidia 630GTX, Harmony Home Control, Pioneer VSX-53, Panasonic VT30 65" 3D TV, Windows 10, Isengard
Bedroom HTPC: Zotac-ID 41 8GB RAM, 128GB SSD, Rii micro keyboard remote, Samsung HW-E550, Sony 32" Google TV, OpenElec 6.0 beta 4
Reply
  • 1
  • 2
  • 3
  • 4(current)
  • 5
  • 6
  • 11

Logout Mark Read Team Forum Stats Members Help
NFS issues continue with Gotham beta 20