Android MiniClient v1.2.0 - Released

[panteragstk]

Originally Posted by stuckless

I can see that the UI is hanging... but beyond that I can't see why. Is server debug logging turned on? Maybe Jeff can spot something.

I did have an issue on unRAID where the drives would spin down while I was watching a show... and then when the show ended, i'd get the spinning circle for a few seconds and then it would return me to the UI. This was because the drive that had the show was spun up, but when the show ended, and I had to get back to the UI, the UI would be accessing data on a drive that was spun down, and I'd had wait for it to spin up, before it return to the UI. I fixed it by telling unRAID to NOT spin down my drives.

I attached the server log to my previous post. I also tell unraid to not spin down my drives.

The other section of the log that was strange was that it seems that the server was looking for something to record on all of my tuners after the stop command was issued. None of the things in the list are actually scheduled recordings. It starts with "looking for potentials".

See attached.

[stuckless]

Originally Posted by panteragstk

Strange thing is that this is new. It didn't happen until the last update to the mini client, but I think the docker got updated as well.

For me... it just happened all of sudden as well... didn't even coincide with an update, etc. Because unRAID started to utilize another disk that it wasn't using before. If you can reproduce, this consistently... you can bring up the unRAID dashboard which shows if a disk is spun down or not, and then see if any are spun down before you hit stop (that's what I did), and sure enough when the disk was spun down, and I hit stop, I'd get the spinning. And then after that it would show spun up.

For me, setting all disks to never spin down solved it... a better solution might actually keep all appdata on the ssd cache.

[Fuzzy]

Originally Posted by stuckless

For me my appdata is NOT on the cache drive, which means that it's spread out over several disks in the pool... When I would stop, gemstone would need to reload the VFS and fetch fanart. Because of the image cache size on the server, the images that were needed needed to re fetched, so that forced a disk access. I don't have a large ssd cache, so I don't store the appdata on the cache (and the cache data is not parity protected).

Not parity protected, but I do believe unRAID supports multiple cache drives, and it will RAID1 them by default (haven't verified, going off memory of forum discussions).

[panteragstk]

Originally Posted by Fuzzy

Not parity protected, but I do believe unRAID supports multiple cache drives, and it will RAID1 them by default (haven't verified, going off memory of forum discussions).

You are correct.

[bialio]

Not related to the stop issue - but on my Windows Server based install I'm starting to get REALLY laggy UI on the mini client - happens about 3 out of 5 times. It sits on the initial 'Connecting to <server>' screen longer than usual, and then UI updates are very slow. Once I start a show video playback is fine - it's just UI rendering that suffers.

I'm running 1.2.6 on the Shield, ethernet connection.

[stuckless]

Originally Posted by bialio

Not related to the stop issue - but on my Windows Server based install I'm starting to get REALLY laggy UI on the mini client - happens about 3 out of 5 times. It sits on the initial 'Connecting to <server>' screen longer than usual, and then UI updates are very slow. Once I start a show video playback is fine - it's just UI rendering that suffers.

I'm running 1.2.6 on the Shield, ethernet connection.

I've had long "connecting" periods as well... but I also got them on the HD300... The period of "connecting" is setting up the server context and prepping images... I haven't really looked to deeply into why it would sometimes take longer to connect.

One issue that did bother me, and why I increased the image caches on the client was, using some of the more fanart intensive UI (gemstone/phoenix) i'd get lags as I navigated items. I discovered this was related to the size of the image cache on the client. I was using a fixed 32mb cache, but some images that were being send across were 20+mb in size. This meant that before the UI would render it.. it would be unloading the "old" images to make room for the new ones. On the server this was taking 1-2 seconds become some of these images are quite large.

In phoenix core, I'm actually rescaling the images as they are downloaded, because there's point in having a 2000x3000 poster when the max size you'll ever show would only be the same as the UI height. Doing this dramaticaly increased the performance of the UI... since sagetv didn't have to load/cache it ont he server and the UI image cache could hold more images as well.

On my Shield, I actually set my UI image cache to be 192mb (default as of the latest release is 64mb, and before that it was 32mb).

I'll try to add more logging information around connecting to see where the connecting pauses are coming from.

[bialio]

Yeah, and I think both Gemstone and Phoenix are caching scaled images so that we aren't loading full size from the server.

But the connecting period being long is usually a precursor to all the UI being slow. I did up all of my image caches to the max as soon as I noticed those settings last week. So I don't think it's just a server context initialization /etc issue. Something else is clogging up the pipeline.

I'll try and catch some logs next time I see it. It happens pretty frequently.
btl.

[stuckless]

Originally Posted by bialio

Yeah, and I think both Gemstone and Phoenix are caching scaled images so that we aren't loading full size from the server.

But the connecting period being long is usually a precursor to all the UI being slow. I did up all of my image caches to the max as soon as I noticed those settings last week. So I don't think it's just a server context initialization /etc issue. Something else is clogging up the pipeline.

I'll try and catch some logs next time I see it. It happens pretty frequently.
btl.

Phoenix does a per-client cache, if I recall, so the images are exactly the size it needs (which is nice). Gemstone uses a shared cache and while images are rescaled, they are not rescaled exactly to what the UI needs, and the final scaling is done in the UI layer (ie, specifying the image size. I found a case in gemstone where in one of the flows it was using the full poster size (2000x3000) or percentage size of that (say 80%), which is still large, so that was eating up lots of image cache on the client and server resulting in UI lags when you navigate.

In the miniclient, the "making room" part consists of "unloading" the image (which sends a request to the server to have it unloaded there as well). This is done as a serial process, so depending on how many images have to be unloaded it can introduce a lag. I'm thinking that it should be able to be done as a background task.. ie, I can get the images that need to be unloaded, to make room for the new image, and then in a background task, tell sagetv to unload them all. It doesn't do that today.

[EnterNoEscape]

Originally Posted by Fuzzy

Not parity protected, but I do believe unRAID supports multiple cache drives, and it will RAID1 them by default (haven't verified, going off memory of forum discussions).

It forces btrfs when you use more than one drive for the cache drive. It implements for lack of better terminology, RAID 1. What is really happening is a copy of every "file" is made to a different drive. You can configure just how many copies are made. Because it is structured like this, you can actually have drives different sizes and it will result in more available space than you might get with traditional RAID 1. Also btrfs hashes the files and because of this knows which copy is the good copy if a drive gets corrupted. According to the unRAID website, it also leverages btrfs with Docker to effectively deduplicate redundancies in containers (do not confuse this with btrfs deduplication programs; it uses the same mechanisms, but only applies to Docker). This is a very high overview and in this description "file" is probably not the best word choice either.

I am actually using 8 SSD drives balanced across two controllers for bandwidth reasons as my cache and the performance is astounding at times. I actually did do performance benchmarks on various parts of my system to determine that the SSD's where being starved of bandwidth. I just happened to have two sets of 4 from other caching related projects and I wanted to do something interesting with them instead of just letting them collect dust.

Something of note is when the cache drive operates many magnitude faster than the actual array, you will run into interesting write cache memory management problems particularly if you're using Dynamix Cache Directories, you need to add the option "-p 100" which is the system default before the plugin changes it to 20 or the cache pressure will be so low that you will have VM's being killed due to OOM errors. I have 32GB of ram and about 18GB of that is being used by VMs. I would sometimes wake up in the morning to find out that a few VM's where killed due to OOM errors while the mover was running. Increasing the cache pressure fixed my problem permanently.

[BobbyDing]

Originally Posted by Fuzzy

It does decode AC3 and DTS, but I think only outputs stereo sound. PLIIx is the sound effects system in the receiver. PCM just means a digital signal coming in - I'm guessing it's 2.0 PCM going in though.

Thanks Fuzzy. Indeed if I select "Direct" on the amp remote the front panel indicators drop back to Stereo displaying "DIG PCM" and all the audio comes out the front L/R channels. If I select "Movie/PLII" the amp lights up "DIG PCM, DD, PLIIxC and S.BACK. I can confirm that in this mode the dialog is only coming out the center channel and the music, effects are coming out the front L/R and rear L/R. I don't know if this happened during a recent IJK upgrade or maybe I had never forced the receiver into this mode before (on my Denon amp this mode doesn't happen automatically, it appears I have to select it. Anyway I thought I'd mention it as I do seem to have surround using IJK. I'm curious if anybody else can confirm this.

Bobby

[bialio]

One suggestion Stuckless regarding the miniclient. I notice that if there is video playing then a long press brings up the built in navigation with pause highlighted. In other scenarios options is highlighted by default. This is nice. Very useful.

The problem is that if you are looking for options while playing back video the navigation path to get there is pretty arduous :

up, left, left, left, left, left, left, down.

Maybe that first move past the Navigation controls row could move over to the options button?

btl.

[Fuzzy]

Originally Posted by BobbyDing

Thanks Fuzzy. Indeed if I select "Direct" on the amp remote the front panel indicators drop back to Stereo displaying "DIG PCM" and all the audio comes out the front L/R channels. If I select "Movie/PLII" the amp lights up "DIG PCM, DD, PLIIxC and S.BACK. I can confirm that in this mode the dialog is only coming out the center channel and the music, effects are coming out the front L/R and rear L/R. I don't know if this happened during a recent IJK upgrade or maybe I had never forced the receiver into this mode before (on my Denon amp this mode doesn't happen automatically, it appears I have to select it. Anyway I thought I'd mention it as I do seem to have surround using IJK. I'm curious if anybody else can confirm this.

Bobby

That is completely the result of the receivers post processing. Dolby Pro Logic will extract sound that is the same in both Left and Right speakers and send that sound to the Center channel, and will send sound that is reversed phase from each other in left and right to the rear speakers. It will then send what remains to left and right. This is the 'OLD' way of broadcasting surround sound over stereo medium.

[KryptoNyte]

I've been using the miniclient on a Galaxy Note 4 for a while now and it works great. A couple releases back, I noticed that [during playback only] there is now a black bar top and bottom of the video, that changes the aspect ratio, so folks look a bit short and fat. The black bar(s) are about the height of the phone's android notification bar, but the notification bar is not displayed unless I manually pull it down during playback.

Is there a way to get rid of the black bars top and bottom and get the aspect ratio back to good?

[bialio]

Originally Posted by Fuzzy

That is completely the result of the receivers post processing. Dolby Pro Logic will extract sound that is the same in both Left and Right speakers and send that sound to the Center channel, and will send sound that is reversed phase from each other in left and right to the rear speakers. It will then send what remains to left and right. This is the 'OLD' way of broadcasting surround sound over stereo medium.

My question is does the player understand Dolby Pro Logic enough to take AC3 and encode it so that the receiver is actually getting a ProLogic stream.

[stuckless]

Originally Posted by bialio

One suggestion Stuckless regarding the miniclient. I notice that if there is video playing then a long press brings up the built in navigation with pause highlighted. In other scenarios options is highlighted by default. This is nice. Very useful.

The problem is that if you are looking for options while playing back video the navigation path to get there is pretty arduous :

up, left, left, left, left, left, left, down.

Maybe that first move past the Navigation controls row could move over to the options button?

btl.

Doesn't the Mic button bring up the options during playback? But yeah, the TV navigation controls could be better as well. Currently other than setting a default button based on the state of the UI (ie, video or no video), I don't actually control the navigation flow of the button highlights... i could... I just havent dont' much around that... Personally, I'm think that maybe we need 2 navigation screens... one for video but a certain button layout and one for UI (that wouldn't need the video controls, etc).

Originally Posted by KryptoNyte

I've been using the miniclient on a Galaxy Note 4 for a while now and it works great. A couple releases back, I noticed that [during playback only] there is now a black bar top and bottom of the video, that changes the aspect ratio, so folks look a bit short and fat. The black bar(s) are about the height of the phone's android notification bar, but the notification bar is not displayed unless I manually pull it down during playback.

Is there a way to get rid of the black bars top and bottom and get the aspect ratio back to good?

Could you post a screenshot?

[BobbyDing]

Originally Posted by Fuzzy

That is completely the result of the receivers post processing. Dolby Pro Logic will extract sound that is the same in both Left and Right speakers and send that sound to the Center channel, and will send sound that is reversed phase from each other in left and right to the rear speakers. It will then send what remains to left and right. This is the 'OLD' way of broadcasting surround sound over stereo medium.

I understand now. Thank You!

Bobby
.

[Fuzzy]

Originally Posted by bialio

My question is does the player understand Dolby Pro Logic enough to take AC3 and encode it so that the receiver is actually getting a ProLogic stream.

IJK is based on ffmpeg, and ffmpeg does have the capability to downmix to dolby surround and dolby pro logic ii (-matrix_encoding dolby or -matrix_encoding dplii)- but that doesn't mean IJK is telling it to do so.

[VCRUser]

My server is Sage 6, and each time I've tried to run this app, I get a barely-functional UI: screen-taps activate nothing at all or something vaguely near what's displaying on screen where I tap. I've accidentally invoked the navigation overlay, which unhelpfully shows transparently over whatever is showing underneath, making it very hard to read, and tapping the various icons on the overlay have no, or random effects. At one point "watch live tv" got invoked and I had the worst time getting it to stop. I had to tap random things to get access to my devices's "back" or "home" button and was about to just shut it off.

Surely that isn't the way it's supposed to work. Since I've had similar behavior on various Android client devices, could it be that the client doesn't like 6 as a server? I keep thinking it would be a really great thing to gave if only I could make it work.

[Fuzzy]

You definitely will want to upgrade to v9 on your server.

[stuckless]

Originally Posted by VCRUser

My server is Sage 6, and each time I've tried to run this app, I get a barely-functional UI: screen-taps activate nothing at all or something vaguely near what's displaying on screen where I tap. I've accidentally invoked the navigation overlay, which unhelpfully shows transparently over whatever is showing underneath, making it very hard to read, and tapping the various icons on the overlay have no, or random effects. At one point "watch live tv" got invoked and I had the worst time getting it to stop. I had to tap random things to get access to my devices's "back" or "home" button and was about to just shut it off.

Surely that isn't the way it's supposed to work. Since I've had similar behavior on various Android client devices, could it be that the client doesn't like 6 as a server? I keep thinking it would be a really great thing to gave if only I could make it work.

Certainly it will run better with a v9 server... but if you are using this on $99 tablets... it's going to perform like a $99 tablet. This is a client that basically emulates the HD300 (for a lack of a better work), so it is meant for a TV. It can run on a Phone/Tablet, but the experience will be sub-par... On decent phones/tablets it'll run "ok"... but if you are running this on cheap/old phones/tablets... then it's going to suck...