Server Version#: 1.29.2.6364, 1.31.1.6733, 1.31.2.6783
I recently updated my server to 1.31, and immediately noticed that scans of new content were taking longer than I expected. After scanning something new, it would sit with the progress wheel looking fully completed, but it would stay that way for 5-10 minutes each time.
As an experiment, I set up several new servers in docker to test. I copied the database 2 times, and installed PMS 1.29.2.6364 on one, and PMS 1.31.1.6733 on the other. Also, my main server (which is running natively on Ubuntu) was updated this morning to 1.31.2.6783.
So just to be clear, all 3 instances of Plex are running on the same hardware. All 3 started with the same database, and all 3 are using the same media.
I use autoscan to trigger scans on Plex, rather than scanning the entire library. So when new content is added, I have it set up to trigger a scan on all 3 Plex servers.
When a scan is triggered by autoscan, the version of PMS running 1.29.2.6364 scans the new media very quickly, and completes within a couple seconds at most. The docker instance running 1.31.1.6733 and the native instance running 1.31.2.6783 both exhibit this issue, and sit for several minutes with the progress wheel almost at 100% before moving on to the next item. If there are several items to scan, this can add up quickly and sometimes it will take a very long time.
Settings - Server - āShow Advancedā - Library - āGenerateā¦*ā ā āAs a scheduled taskā (only)
I have all of those set to āNeverā on all of the serversā¦
By the way, I have checked all the settings to make sure they are all the same on all servers.
āGenerate credits video markersā is a new capability that is enabled by default in both the Settings | Library and also as āEnable credits detectionā in the Advanced section of each video library.
Disable those, restart Plex and see what happens.
Fully disabled on my system and Iāve the same behaviour. Another variable is the new audio and subtitles code.
Yep, disabled on mine as well (see screenshot posted above).
I just did an experiment using a full scan (something I donāt normally do since I use autoscan).
I ran a manual scan of the entire library, first with the PMS I normally use, and secondly with one of the extra docker installations I have installed for testing purposes.
The instance running PMS 1.29.2.6364 took 4:39 to run a full scan.
The instance running PMS 1.31.2.6783 took 13:52 to run a full scan.
Again, I want to reiterate, these are identical databases (in terms of content) and they both point to the same media.
Also, I want to point out that there was no new content added during these scans. Everything had already been previously scanned via an autoscan scan.
So basically the only difference between these two is the version of PMS, and the newer version took more than 3x longer to run the scan.
If Iām watching the UI during the scans, most of this extra time is taken at the end of each library, at least according to the progress indicator. It will travel clockwise all the way around while the scan is happening. It then sits for several minutes at the end of the cycle, before moving on to the next library in the database. By comparison, the 1.29.2.6364 instance immediately moves on to the next library as soon as the progress indicator finishes the cycle.
1 Like
If you had the time would you mind doing the following ?
On the system with 1.29.2.6364 which runs the what Iāll call normal speed scan could you wait 30 seconds and then run it again without doing anything else and report the results ?
Reason I ask is from my own thread which I linked above this slow scan issue only seems to happen after a successful normal speed scan.
Also, what OS distro, version and kernel are you running ? Iāve another theory Iād like to rule in or out based on someone else with the same slow scan issue. Thanks.
I actually donāt think these are the same issue, to be honest. They might have similar symptoms. but they donāt match up exactly.
I just did what you asked: ran the same full scan on the 1.29.2.6364 instance, 30 seconds apart.
Pass 1 : 4:31
Pass 2 : 4:52
On the older PMS software, the full scan is consistently around the 4-5 minute mark. With the newer software, it is always more than 3x that long.
1 Like
Okay, thanks. Letās see if someone from plex comes along to add anything.
I also want to point out, my issue doesnāt only apply to full scans. I have a video I would like to share, but the forum wonāt let me upload it. Itās a screen capture of the activity, showing two servers importing the same individual episode at the same time. The 1.29.2.6364 instance imports the episode in seconds, then gets out of the way. The new PMS software imports the episode, then sits with the progress meter completely fully around, for about 2 minutes, before it finally completes.
Why would anyone from Plex jump in the pool with you 
LOL
Iām in the middle of testing the new releases now (Video transcoding fixes).
When thatās complete, Iāll look at the scanner.
I knew you would see this thread and think, oh great, this guy again.
I appreciate itā¦ Iām trying some other tests on my end as well, to try and get some additional data points.
1 Like
@anon5074910 by any chance do you use Plex Meta Manager to create overlays and/or collections?
Nope, just pure plex server with no plug ins or add ons.
@Departed69
I have your test data. Which should I scan?
I did the scan of the āMoviesā (48761 titles) and it completed in less than 1 minute on my xeon.
Thatās actually terribly outdated at this point. After I did the rescan of all my media, I abandoned that database.
I canāt remember if I had the Collections set up in that database. Can you check and see if there are a bunch of Collections or not?
Remember, I have just the raw names. I am not using your ādatabaseā per se.
I made dummy files from extracting the pathnames found in that old DB.
At present count, your + mine + test files = 801K files.
Iām letting it finish scanning in (i needed to add one section).
When that is complete,
- Scan the entire server
- Get count of files indexed (and recognized)
- Do the math to determine: Items/second -or- Seconds/item.
Make sense?
Iām not sure I understand the purpose or outcome, but as long as you have a planā¦
The reason I was asking about the collections is, I originally had a bunch of collections set up through Plex Meta Manager. I realized that one of the differences between the two (the one that takes time and the one that doesnāt) is that the one that takes a long time to scan still had those collections. After I deleted the collections, the scan time improved drastically.
Of course I donāt know if this correlation makes sense or not, but hopefully the presence of collections wouldnāt cause the scan to take an inordinate amount of time. But thatās why I was asking if your copy of my data included the collections or not.
My plan is to recreate the collections on the ātestā versions of PMS running in docker. I now have two test beds in docker, one running 1.29.2.6364 and another running 1.31.1.6782. Once I run PMM against both of those and generate the collections, I can determine whether one version of PMS works better than the other under those conditions.
