34 replies to this topic

[BadBlake]

I have 2 Vu+ Duo 2 i my house running OpenPli 8.1, and and both are mounted remotely so that I can watch recordings I've made from both of them.
The problem is that when I access the movie folder on my remote box (from either one of them)it takes forever in a day to read what's in the folder. The update is so slow. It takes about 15 seconds to update each recording visible in the list.
Ater about 30 min the list is completely visible. When trying to view for example the last movie on this list -I get a black screen -nothing happens. I press stop and it returns me to the movie list. If I the choose the first recording on the list -it plays properly.
There's something very wrong regarding the update process. Same result from both boxes and my Lan speed is excellent. Both boxes are wired too.

[BadBlake]

I have 2 Vu+ Duo 2 i my house running OpenPli 8.1, and and both are mounted remotely so that I can watch recordings I've made from both of them.
The problem is that when I access the movie folder on my remote box (from either one of them)it takes forever in a day to read what's in the folder. The update is so slow. It takes about 15 seconds to update each recording visible in the list.
Ater about 30 min the list is completely visible. When trying to view for example the last movie on this list -I get a black screen -nothing happens. I press stop and it returns me to the movie list. If I the choose the first recording on the list -it plays properly.
There's something very wrong regarding the update process. Same result from both boxes and my Lan speed is excellent. Both boxes are wired too.

One more thing -I also mounted the hdd on my win7 computer to both boxes, and it works perfectly. No update lag at all.

[WanWizard]

Depending on the configuration it needs to access every recording to determine its run length. This may take a while on a slow link.

 

Also, it reads the .eit and .cuts file to display the programme information of the recording, and the indicator of how far you've watched it.

 

You can configure both in the expert settings, record menu. But I doubt it will make much of a difference, the Duo2 ethernet port is known to be slow.

 

Also, using CIFS is quite a bit slower than using NFS (you need to install the nfs server package for this, and configure the correct location to export).

[BadBlake]

Depending on the configuration it needs to access every recording to determine its run length. This may take a while on a slow link.

 

Also, it reads the .eit and .cuts file to display the programme information of the recording, and the indicator of how far you've watched it.

 

You can configure both in the expert settings, record menu. But I doubt it will make much of a difference, the Duo2 ethernet port is known to be slow.

 

Also, using CIFS is quite a bit slower than using NFS (you need to install the nfs server package for this, and configure the correct location to export).

I've never had this problem before with other images like Open ATV, blackhole og even Openpli 4 or 5 images, and why don't I have this problem viewing files located on my Win7 computer. Also Cifs share?

[WanWizard]

From both Duo2's ?

[BadBlake]

From both Duo2's ?

Yes

[BadBlake]

 

From both Duo2's ?

Yes

 

I can't even view screeshots  on my remote boxes -that's how poor this is working

[40H3X]

Hmm maybe something wrong with NIC on 1 of your Duo2's, half duplex maybe?

 

Maybe check it and do a

ethtool eth0

 

[BadBlake]

 

Hmm maybe something wrong with NIC on 1 of your Duo2's, half duplex maybe?

 

Maybe check it and do a

ethtool eth0

No, that's not it. As I mentioned I have no problems viewing files on my mounted Win 7 computer -from both boxes. Also as I metioned above -never had this problem with earlier images of pli or other images that I've tried in the past.

[WanWizard]

None of that is very relevant. And dismissing any questions for debugging doesn't get you to a solution quicker.

 

Both the PHY and the drivers of the Duo2 are notorious, particularly the lack of memory in combination with the lack of scatter/gather functionality. The last is specific to ingress traffic, while accessing files from a Win7 computer is egress traffic and not impacted. The newer your OpenPLi version, the less memory you have available.

 

What rsize and wsize have you used for the mounts on both boxes? Does reducing both to 8192 help?

[40H3X]

......and if I may add, how do you know when you have not checked it. If there was some general problem with CIFS, it would be raining complaints here.

[ccs]

I'd use NFS for the 2 Vu's to access each other, and CIFS to get to them from windows.

[BadBlake]

None of that is very relevant. And dismissing any questions for debugging doesn't get you to a solution quicker.

 

Both the PHY and the drivers of the Duo2 are notorious, particularly the lack of memory in combination with the lack of scatter/gather functionality. The last is specific to ingress traffic, while accessing files from a Win7 computer is egress traffic and not impacted. The newer your OpenPLi version, the less memory you have available.

 

What rsize and wsize have you used for the mounts on both boxes? Does reducing both to 8192 help?

I'm a newbie concerning things like this. How to find out what r or wsize i use for mounts, and how to change them? Its probably a memory issue

[WanWizard]

if you type the command "mount", you can see which rsize and wsize is used. If not specified, they are negotiated between server and client, and they might be too large.

 

And that is a memory issue, it needs continuous memory blocks this size to be able to buffer the traffic, the larger the sizes, the larger the buffers need to be, the earlier you run into memory issues.

 

You can alter the sizes of the mount in the mount options, disconnect the mount, add "rsize=8192,wsize=8192" to the mount options, and mount again. Use the "mount" command to verify".

[Beanie]

I think I am seeing the same problem.

 

VuSolo2 with internal HDD

ZGemma H9.2H with external USB HDD

VuZero

 

All 3 are connected to ports on the same gigabit switch, which reports full duplex at 100Mb or 1000Mb, depending on the NIC in each box. The physical connection rate does not change

 

The HDDs in the ZGemma and VuSolo2 are mounted using CIFS on the VuZero. Using OpenPLi 7.3, this worked perfectly. Recordings could be browsed and played normally. When I upgraded to OpenPLi 8.0, and also 8.1, opening the file list (movie folder with about 20 items) is very slow. It takes about 3 minutes to show the file list, with each entry taking about 10-15 seconds to show on the screen.

 

I have tried many different versions of OpenPLi on each box, and this is what I found.

It does not matter which version of OpenPLi (7.3.x, 8.0.x, 8.1.x) is the CIFS client box.

It does matter which version of OpenPLi is the server (samba). With 7.3.x on the VuSolo2 or ZGemma, CIFS/samba shares work well. With OpenPLi 8.0.x or 8.1.x as the server, then the problem is there.

 

I have tried with a clean flashed image (no settings carried across) and that does not change anything.

 

The only difference that I could see was a change in the version of Samba included with OpenPLi 8.0.x and above. The samba config files are identical. I tried reading the changelogs for Samba, but there are so many changes, I do not know what it may be.

 

I do not think this is a Vu+ problem, as the ZGemma behaves exactly the same.

[WanWizard]

The same reply applies as above.

 

Check the rsize and wsize on the mount on the client box, and see if making it smaller helps.

 

The changelogs of Samba are not relevant, as you say it doesn't matter what the server side is. The client side CIFS filesystem drivers aren't part of Samba, they are part of the linux kernel, and haven't changed (unless you have an Edision, the only vendor that updates their linux kernel regularly).

 

If I make a mount to my NAS using the mount manager, it creates:

//172.16.1.31/OpenPLi on /media/net/movies type cifs (rw,relatime,vers=3.0,sec=ntlmssp,cache=strict,username=openpli,domain=NAS03,uid=0,noforceuid,gid=0,noforcegid,addr=172.16.1.31,file_mode=0755,dir_mode=0755,iocharset=utf8,nounix,serverino,mapposix,rsize=1048576,wsize=1048576,actimeo=1)

which uses both read and write buffers of 1 MB.

 

Which is way to large for all but the most big and powerful boxes with loads of memory. Reduce it to 32 KB or even 8 KB on both VU's. The H9 should be capable of using bigger size buffers (can't test that myself, my H9 is connected via wireless).

 

Write and read speed on the above CIFS mount (my Duo4K -> my NAS):

root@vuduo4k:~# dd if=/dev/zero of=/media/net/movies/testfile1 bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.0GB) copied, 11.252369 seconds, 91.0MB/s

root@vuduo4k:~# dd if=/media/net/movies/testfile1 of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.0GB) copied, 11.822706 seconds, 86.6MB/s

I use NFS everywhere, that same mount using NFSv4:

root@vuduo4k:~# dd if=/dev/zero of=/media/nas/testfile2 bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.0GB) copied, 12.876373 seconds, 79.5MB/s

root@vuduo4k:~# dd if=/media/nas/testfile2 of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.0GB) copied, 10.354145 seconds, 98.9MB/s

which in both cases is close enough to Gbps line speed.

[Beanie]

Thank you @WanWizard.

 

Here is the relevant output from mount on the client machine (VuZero, currently running OpenPLi 7.3)

//192.168.1.60/Harddisk on /media/net/VUSOLO2 type cifs (rw,noatime,vers=2.1,sec=ntlmssp,cache=strict,username=guest,domain=VUSOLO2,uid=0,noforceuid,gid=0,noforcegid,addr=192.168.1.60,file_mode=0755,dir_mode=0755,nounix,serverino,nobrl,rsize=65536,wsize=65536,actimeo=1)
//192.168.1.76/usb on /media/net/ZGemmaH9 type cifs (rw,noatime,vers=2.1,sec=ntlmssp,cache=strict,username=guest,domain=H9,uid=0,noforceuid,gid=0,noforcegid,addr=192.168.1.76,file_mode=0755,dir_mode=0755,nounix,serverino,nobrl,rsize=65536,wsize=65536,actimeo=1)

Both mounts created using mount manager. The VuSolo2 is running OP 7.3. H9 is running OP 8.1.

Browsing and playing movies from the VuSolo2 work perfectly, from the H9 the problem exists. Both mounts have the same rsize & wsize. The only difference is the OpenPLi (including Samba) version at the server end.

 

Upgrading the VuSolo2 from 7.3 to 8.1 causes that mount to be problematic as well.

 

Dropping rsize & wsize from 64k down to 8192 helps a little, but does not solve the problem. It improved how quickly the file list appeared on screen, but some files are showing without the creation date, some won't play and some will only play the first few minutes of the file.

 

Changing the VuZero client OpenPLi version from 7.3 to 8.0 or 8.1 makes no difference.
 

[WanWizard]

No clue, and as you can see, I can't reproduce it (all my boxes run the current develop), but I don't have a single box with an internal disk.

[Beanie]

Doing more troubleshooting, I cannot see anything obvious in the samba logs or smb status between 7.3 and 8.0 / 8.1, other than having a later samba version. smbstatus and the logs (at log level 2) seem to be the same between devices.

 

The switch is reporting much higher throughput than expected when connecting to the H9 (OpenPLi 8.1) share though. Traffic sitting at about 60Mb/s when trying to watch a recording that is about 4Mb/s bitrate (and failing to watch it). Network throughput is consistent with recording bitrate when watching something from the OpenPLi 7.3 share.

[WanWizard]

That  is very odd...  I'll see if I can reproduce that.