Progressively slower media start time on Plex

When I first setup rclone w/ crypt on ACD about 2 months ago, performance was so good it was close to my local NAS.

Instead of 2-3 seconds to start a movie/show it would take 4-6 seconds but would plex flawlessly. This is 4-8GB/hr TV shows and 25-35GB/movies.

Over the last month in a half, it has gotten worse and worse. Now it takes 1-3 minutes to start a show. Sometimes it stutters at the start, if I stop it and restart it usually resolves itself. The stuttering is rare, but the start time is consistent.

I only have 3.5TB on ACD. I turned off thumbnail creation in the scanner which caused a lot of slow downs early on. I know the scanner for the cloud version of plex was optimized for amazon (even though they abandoned it) but as far as I know there is no way to optimize it further.

I do run no encoding stream (original setting) with high bitrate remuxes but I do not typically run multiple streams, and it is running on a 2xquad core Xeon 5560 w/ 48Gb ram.

This is my mount command:

sudo rclone mount --read-only --allow-other --max-read-ahead=2G --verbose --acd-templink-threshold=0 --contimeout=15s --checkers=20 --bwlimit=0 --retries=3 --timeout=30s --low-level-retries=1 --transfers=8 --stats 0 --dir-cache-time=5m --log-file /tmp/rclone-acd.log ACDE: /mnt/ACD &

Something is off with latest rclone i noticed the slower performance as well, I suspected ACD hence i made that single file copy script so i can test performance.

As for mount you dont need all those flags, this is my current setup.

rclone mount
–acd-templink-threshold 0
–stats 0
acdcrypt: /storage/acdcrypt/ &

Dont change default retries and timeouts with even lower values as that only mean stream will fail.
–max-read-ahead=2G its pointless since 128K is max by OS and default in rclone
–transfers 8 - why limit anything since rclone will only use 1 transfer per open file
–dir-cache … 5 min is default
–verbouse & --log-file you can get slightly better performance without it. ( use it only when you plan to debug and then add fuse debug and drop headers there as well)

To make sure you have good peering with amazon test single file copy test
( and copy from your mnt/ACD not acde directly same way as Plex would access files )

I recently switched from ENCFS to CRYPT and now with over a week of testing performance is worse and i can put my finger on it why, since i was doing some partial tests and they were always on pair. Even when I switched on day 1 I spend 1h of testing ECFS playback time and then switched to CRYPT and repeated all the tests.
( it worked more or less the same )

@ncw is it possible that the longer crypt mount is on the worse performance get ( highly unlikely but cant think of anything else ) … need to restart server and do some more testing. ( iam usually not restarting server at all )

Just an FYI but I’ll reiterate that I’ve had my ACD encrypted mounted for 10 days and Plex has been rock solid. I’m on version v1.35-42-g9fdeb82

You really need to identify which part of the chain (ACD to Server, Transcode, Server to client) is causing the issue and troubleshoot from there.

I’ve used PlexPy so much recently with troubleshooting it really is great. For example if the Plex buffer is at 20% and my progress is at 5%, I know it’s not the connection to the server from ACD. If my buffer can’t fill up or the transcode speed (hit the little “i” icon in PlexPy now playing) is 0.0 or less than 1, I know it’s an issue either with ACD to Server or the transcode speed (disk issue, IO issue, network issue, etc.).

I also have Plex debug on and use tail -f Plex\ Media\ Server.log and look for the telltale signs of transcoding, ie:

We want 30 segments ahead, current max segment 30 (that’s a problem if they’re the same), you may also see -1 there, thats also a problem.

Disabling Direct Play and Direct Stream can help you as well. DirectPlay and DirectStream for as great as they can be are extremely volatile to use in a Plex cloud server setup because everything has to be perfect to ensure a smooth experience.

With transcoding, if you can pull down enough data and configure Plex to buffer more in the Transcoder settings) you can suffer a few glitches in the ACD to Server or Server to client chain without impacting the experience.

All things I’ve been delving into while troubleshooting inconsistent issues.

How do you use Plexpy? It doesn’t seem to do much different than just pulling the playing media from activity and hitting the (i) outside of accessing logs.

More importantly, playback is generally fine. It’s starting, and plexpy does not show anything playing until it actually gets past the spinning loading wheel. That’s my big problem, that wheel is there for 60-180 seconds. Once it finally starts, 98% of the time it is flawless (as long as I don’t try to skip forward/back).

I’m not limiting transfers, I’m actually doubling default (4) to 8. Also increased checkers so it starts with more.

Checkers and transfers do NOTHING for the mount command. Ncw confirmed this.