Advice on performance of new install

[powderburned]

I could use some advice on performance tuning with F23 and A9. I was using A7 on Fedora previously and performed a fresh install of the latest versions of the software this week. After considerable issues with the installation (see post viewtopic.php?f=4&t=6499) I was eventually able to get everything going "out of the box."

Currently I am seeing significantly reduced performance when playing videos in Kodi media center on a wired gigbit ethernet home network as compared with the same hardware running amahi 7 "out of the box." Videos that once played without delay or buffering now require 5-10 seconds to access and frequent(a pause every few seconds to few minutes depending on video) pauses for buffering. Most of the video content is mpeg4 720 videos but I have noticed that the larger the video the worse the issue becomes.

I made the changes outlined in your article on samba tuning but only saw a marginal improvement. What I don't understand is that single file transfer rates between the hda and the media center pc are about the same as they were before the upgrade (as rated by the windows file transfer interface). I am noticing some significant latency resolving the hda and then opening the shares (1 to 2 seconds to open a share).

I am not seeing much load on the processor at all. Memory usage for kernal and apps is low but memory caching is consuming remaining available RAM. (4GB total). However, caching takes some time to fill remaining ram space and performance is no better during the period when there is plenty of unallocated space available.

Below is pretty typical.

Code: Select all

Type Usage Free Used Size Physical Memory 98% 87.21 MiB 3.77 GiB 3.85 GiB Kernel + applications 15% 597.87 MiB Cached 77% 2.97 GiB Buffers 6% 219.77 MiB Disk Swap 0% 5.00 GiB 0 B 5.00 GiB

Please let me know what info would be helpful.
http://paste.fedoraproject.org/373793/88388514

[powderburned]

I am seeing the following error over and over in nmbd.log files:

Code: Select all

[2016/06/06 10:38:38.301944, 0] ../source3/nmbd/nmbd_browsesync.c:354(find_domain_master_name_query_fail) find_domain_master_name_query_fail: Unable to find the Domain Master Browser name WORKGROUP<1b> for the workgroup WORKGROUP. Unable to sync browse lists in this workgroup. [2016/06/06 10:53:41.058911, 0] ../source3/nmbd/nmbd_browsesync.c:354(find_domain_master_name_query_fail) find_domain_master_name_query_fail: Unable to find the Domain Master Browser name WORKGROUP<1b> for the workgroup WORKGROUP. Unable to sync browse lists in this workgroup.

The only error I see in smbd.log is a reference to being unable to connect to the mysql server. I think this is due to restarting the server in whole and the service not being started yet but I am not certain. I think this is the case because I only see this error on a full server restart. when I restart the samba service, I do not see this error.

I do not see any other errors generated and have included the logs from a fresh restart today. I also reverted my smb.conf back to original configuration before the restart. No change in behavior since restart.

[powderburned]

I wound up going back to Amahi7 on F19. All of the metrics in Monitorix look the same but the issues are completely gone. File transfers show the same speed for individual files but a group of files can be read in a directory much faster. For example a sync scan with free file scanner from one of my shares that contains 31K files and folders took 3 minutes to complete on version 9. The same scan on Amahi 7 takes about 30 seconds. At this point I have no idea what's causing the significant slowdown in Amahi 9.

If you have some suggestions, I'm all ears but at this point, I'm staying on version 7.

[cpg]

I don't have an answer. There is no smoking gun that I can see.

With identical hardware and network, I would say possibly drivers?

In the stuff you list, there are no drivers or device information (the fpaste --sysinfo was not done as root?), but still, it does not look like an ancient system and all should be supported.

The mysql error is related as you say to not being ready before smb is up, but once mysql is up, it should be fine. It's for user auth, so it's not really performance sensitive, as it's not in the critical path in data transfer. The other NMB errors are annoying as heck, but they are normal. It's basically "leader election" by searching for other servers in the network dynamically. We asked the samba team to chill them in IRC after investigation, but there was no immediate remedy for that, iirc.

We very rarely see a situation like this, but we have seen it. Someone has bad samba performance like that for some inexplicable reason. We rarely get to get our hands on it to debug further.

If I think of something or I see something else, I will post here. It's definitely unusual.

[powderburned]

I thought about it being related to nic driver but for individual files, I am seeing good throughput. The pause seems to be in accessing and reading the share.

Also, it does appear that there is more network chatter at both idle and share access over the eth0 port in F23.

I'm happy to reinstall 9 and get you whatever logs and info you think is relevant. It will have to wait till after the 20th. Just let me know what to provide and I'll grab it! If you let me know ahead of time, then I can grab the info from both the current working installation and the A9 installation for comparison.

Thank you!

[cpg]

If the issues only come up with large transfers, maybe there is some sort of jumbo frame support issue with your hardware that cropped up in recent releases? Only thing that comes to mind.

[bdaz]

Just stumbled on your post, and I'm seeing the EXACT same problems. To the point where I am about to install FreeNAS and abandon the Amahi platform.

I ran A7 for quite some time with no problems, then upgraded to Ubuntu when it was "supported". I moved back to Fedora after Ubuntu was abandoned about a year ago and have had nothing but performance issues since. Plex and Kodi stutter like crazy, and the rest of the network hums fine.. I've troubleshot network cabling, routers, monitored everything on the Amahi box.. I just can't figure it out.

If you're seeing better performance on A7 then that's what I'll try as the task of moving 6 3TB drives into a different file system (e.g. ZFS) out of greyhole is something i don't really want to spend the time doing ..

[cpg]

Sorry to hear. It does point to hardware support issues, though.

Have you checked MTU settings in your network?

Though MTU being too large would probably lead to hangs, in my experience. At least with SSH. Not sure if Samba or others would automatically retry.

We're testing on Fedora 25, with newer kernel, drivers, etc.
Maybe you can update to Fedora 25 and try? We will support F25 in our next release.

If it's not too current, maybe your hardware is not well supported in Linux?

Maybe you can specify your hardware?

[bdaz]

My hardware is definitely not current (upgraded from an WHS to Amahi 7, to Ubuntu Amahi to A9..) but it worked so swimmingly until the most recent version (I upgraded in 9/2015). It just doesn't seem to deal well with larger files .. I've had to tweak my HTPC quite a bit to buffer a ton of the video to ensure it plays smoothly.. I can't figure out why files seem to be retrieve so slowly from the HDA - and it's only the HDA.

System specs:
CPU: E5200 Intel dual core @ 2.5Ghz (circa 2008)
RAM: 3GB @ 667mhz
NIC: Qualcomm Atheros AR8121/AR8113/AR8114 Gigabit

I'm open to any suggestions. If i could solve this perf issue I'd be happy, but I've had so much frustration since moving to Amahi 9.

Side note: How do i get my hands on Amahi7 to downgrade if that's the best alternative? Re-install from Amahi Express ISO?

[cpg]

I have a few years of professionally working on diagnosing issues in PC systems. I would start by isolating components. Start by not using the network -- use ssh and copy things locally via rsync or cp and see if it replicates.

Then if that works maybe start my mounting a share locally and copying from it. Then *to* it. etc. etc.