November 22, 20178 yr In my current setup my appdata and dockers are all stored and run on my SSD cache drive. I use Plex a lot and having it on an SSD increases the load time for title covers and metadata. I also have SAB that runs and downloads onto the cache drive but then transfers to the array when the download is finished. The issue occurs when SAB downloads a file around 15-20GB+. When it is unpacking/repairing/verifying, Plex becomes completely unresponsive and pretty much refuses to do anything until the SAB file is done and moved and no longer being used. I am wondering what steps I can take to solve this issue. Another computer/unraid server downloading/unpacking/repairing/verifying on its own SSD then use unassigned devices to transfer the file to its destination on the main server? Have SAB download directly onto the array? I just want to find the best solution without losing and download speed, having Plex hang, or losing Plex load time speeds.
November 22, 20178 yr Could be multiple things: - CPU bound on the unpacking - Bandwidth being maxed out on the drives / controllers - SSD problems What I would personally do is basically slowdown the unpacking from SAB by limiting the resources to it (see the docker FAQ) and see if that helps out at all. You could also leave the files on the cache drive, and then let mover put it onto the array when you're asleep.
November 22, 20178 yr Author 6 minutes ago, Squid said: Could be multiple things: - CPU bound on the unpacking - Bandwidth being maxed out on the drives / controllers - SSD problems What I would personally do is basically slowdown the unpacking from SAB by limiting the resources to it (see the docker FAQ) and see if that helps out at all. You could also leave the files on the cache drive, and then let mover put it onto the array when you're asleep. Sorry, I forgot to mention my server specs. I have 2 E5-2650's, 96GB of RAM, and while watching all of this go down in the SAB docker, I have the stats plugin and watch my CPU usage move from 5-10% and occasionally spiking to 20%. My RAM constantly does this: Network and storage usually look like this: Sometimes I have this issue with smaller files causing the file system to be fully allocated but I solve that by balancing the SSD and then trimming it. But that is because of a buildup of files I guess. I don't know how to solve the issue with large files. Possibly with a SSD larger than 500GB? And I don't believe the bandwidth is the issue for the drives since the large file actually sits on the SSD until the mover is activated at around 4AM each day.
November 22, 20178 yr Author I also just found this Glances docker to help me see what is being used. A couple things do not make sense to me. It says my CPU usage is 5.3% but in the center it says SAB is using 60% of CPU. I guess that could mean that it is using 60% of the 5%. But the other thing that sticks out is that I have come warnings about my iowait. This looks like it could be the issue and points toward what you said about the bandwidth/CPU bound issue? Any new idea with this information or does it look like limiting the dockers resources is still your best suggestion?
November 23, 20178 yr Quote %CPU -- CPU Usage : The percentage of your CPU that is being used by the process. By default, top displays this as a percentage of a single CPU. On multi-core systems, you can have percentages that are greater than 100%. For example, if 3 cores are at 60% use, top will show a CPU use of 180%. IE: Since it looks like you have a dual processor system with 8 cores (16 threads) each, your CPU percentages on the task list are on a scale from 0-3200% So your 60.6% actually means 1.9% in use by SAB I've never bothered looking at CPU%. I think that the load averages are a better metric. In your case, a load average < 32 means that there is still cpu resources available. Personally, since Plex is your #1 concern, I would treat it no differently than trying to tune your system for VMs. Limit it to use x number of cores, and limit everything else to specifically exclude those cores. But, it all may be the high IO happening when Sab is unpacking, and Plex is storing its temporary transcode files all on the same drive. A second SSD (mounted via Unassigned Devices) would help by having all of the downloads / unpacking happening on it instead of on the cache drive where the docker.img is stored. (Alternatively, you could also have plex store its temp transcode files in RAM) Edited November 23, 20178 yr by Squid
Archived
This topic is now archived and is closed to further replies.