Recently migrated to a new server, intentionally overpowered it - i9 12900k, 128gb DDR5, 5x 2TB 980 Pro NVMes where the appdata lives - transcoding and operation is buttery smooth, but all scanner tasks are just horribly slow. Taking over an hour per movie to generate preview thumbnails, 3-4hrs to detect intros, sonic analysis is the same story. My host CPU is showing <5% utilization, and PMS has access to the full CPU. Things I have tried:
Several restarts and forced upgrades
iGPU hardware access configured and working
Ramdisk for caching
Disabling ârun scanner at lower priorityâ
Expanding the scheduled task window to 23hrs
Removing and re-adding media, to try to force the scanning/analysis
Shutting down all other processes/containers on the host
The âhiddenâ preferences, forcing âlong running job threadsâ to 20 of my available 24
More reboots and days of waiting
Looking through the logs with debug enabled, I donât see anything wrong that stands out - and it doesnât appear to be âhangingâ on any specific files, because eventually the thing it is working through always finishes. What can I do to open access to the full horsepower of my server?
I donât want to guess, but I want to know the answer because Intro detection of a single 4K TV episode takes at most 5min on a 2014 Celeron inside my old QNAP NAS with no video card.
We probably need a lot more details because nobody else commented:
reboot
demonstrate an intro scan of a single 720p (I just hit analyze to kick it off).
zip and upload server logs, any one that changed after the reboot.
Gory descriptions of your docker could help others. Iâm not familiar enough to know how HW and drivers work on it.
My last server was older, but a dual xeon e5-2670v2, and it also cruised through these processes. Plex was installed in an Ubuntu VM alone in the prior deploy. My new deployment is in unraid. I am very familiar with docker and have run a docker environment for years; just not for plex.
I have reboot a few times over the last week. I just kicked off a new âanalyzeâ and at the moment, generating preview thumbnails appears to be running at an expected rate (2-3min) for the first time since deploying the environment (though my cpu is still around 15%). I have changed nothing at all since last night; as of around 2am, the server was sitting on âDetecting Introsâ for one season of one show for around 6 hours. I just added a new series for the sake of testing, and the same process took around 45 seconds.
When glancing through logs over the last week, the only thing that stood out was the server performing âdeep analysisâ on each video file. I found this while trying to understand why sonic analysis was taking ~1h+ per album. This deep analysis on each video file was taking almost no processing power, and was not reflected anywhere in the interface. Iâm wondering if a library migration forces this as a back end process that is required before the server will execute any secondary processes (like generating previews, detecting intros). Just speculating, but itâs all I can come up with.
Nothing bothers me more than things randomly working after asking for assistance, but here we are - I will continue to watch this and possibly reboot this afternoon to see if the issue returns, or if the issue was truly just related to ânew server break-inâ. Thank you nibbles for your reply and hopefully someone else migrating servers can find something useful here assuming the issue does not return.
Itâs great to hear it working better. I think youâre on to something how a migration or library wide process could slow an otherwise normal system. Plex made an effort to better indicate those in the gui, but I know how debugging this sort of thing can be obvious yet a mystery. I once had a job interview with Alcatel where they threw me a softball question and asked my why some engineering task wouldnât complete but it took me till after lunch to answer itâs probably a hung process.
Ugh. All I needed was to answer top,
and I did not get that job
