Runs used to take an hour or so and are now taking between 5 and 6 hours. I haven't touched the configs or anything else, so just wondering why this might be.
#Kometa runs are taking significantly longer than they used to
worldly cradle
feral copperBOT
Welcome @worldly cradle!
Thanks for being a Kometa Sponsor, we greatly appreciate it! Your ticket will now be diverted to <@&1097919568334311495> and <@&938443185347244033>.
Including meta.log from the beginning is a huge help. Type !logs for more information.
After attaching your log, do not forget to hit the green check boxes when prompted by our bot.
You can press the "Close Post" button above or type /close at any time to close this post.
feral copperBOT
You can press the "Close Post" button above or type `/close` at any time to clos...
worldly cradle
feral copperBOT
feral copperBOT
deft shadow
you're spending almost 5 hours on overlays
episode overlays get expensive. kometa has to look at every epidosde to see if there are changes
worldly cradle
I understand that. The time it takes is new though. It used to be 3-4x faster.
Is it accessing the filesystem where my media resides during this time, or is it getting all of the info it needs about the individual files from the Plex DB?
barren mist
Of:
Run Time: 5:55:59
TV Show Overlays consume:
| Library Overlay Files | 4:54:34 |
Kometa does not access the media files directly; everything is done through the Plex API.
worldly cradle
Gotcha.
And...I'm aware where the time is going. What I'm saying is that this is new behavior.
barren mist
There have been no changes in this area for years. The log I'm looking at doesn't look out of order for 17000 episode-level overlays.
barren mist
Do you have a log from when this very same config took two hours?
barren mist
Just to check, I've got this running and will report back:
#!/bin/bash
test_run() {
docker pull kometateam/$1
docker run --rm -it -v "/home/chaz/kometa/config:/config" kometateam/$1 --run --config config/ackthbpt-config-reset.yml -dc
docker run --rm -it -v "/home/chaz/kometa/config:/config" kometateam/$1 --run --config config/ackthbpt-config.yml
cp /home/chaz/kometa/config/logs/meta.log "/home/chaz/timing-tests/$1.log"
}
test_run "kometa"
test_run "kometa:develop"
test_run "kometa:nightly"
test_run "kometa:latest"
test_run "kometa:v2.2.0"
test_run "kometa:v2.1.0"
test_run "kometa:v2.0.2"
test_run "kometa:v2.0.1"
test_run "kometa:v2.0.0"
test_run "kometa:v1.21.1"
against these two libraries. Not enormous, but they should show any relative timing differences.
I might have to alter the config a bit for the last one there. Will see when it gets that far.
worldly cradle
Here's the exact same library that I just ran.
worldly cradle
My Kometa install runs in a VM that runs on a Proxmox cluster. That VM has 8GM of RAM, and 1 socket with 8 cores.
The one that runs for 6 hours is running on a node that has this CPU:
vendor_id : GenuineIntel
cpu family : 6
model : 154
model name : 12th Gen Intel(R) Core(TM) i9-12900H
stepping : 3
microcode : 0x435
cpu MHz : 400.000
cache size : 24576 KB
physical id : 0
siblings : 20
core id : 31
cpu cores : 14
apicid : 62
initial apicid : 62
fpu : yes
fpu_exception : yes
cpuid level : 32
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect user_shstk avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req hfi vnmi umip pku ospke waitpkg gfni vaes vpclmulqdq tme rdpid movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr ibt flush_l1d arch_capabilities
vmx flags : vnmi preemption_timer posted_intr invvpid ept_x_only ept_ad ept_1gb flexpriority apicv tsc_offset vtpr mtf vapic ept vpid unrestricted_guest vapic_reg vid ple shadow_vmcs ept_mode_based_exec tsc_scaling usr_wait_pause
bugs : spectre_v1 spectre_v2 spec_store_bypass swapgs eibrs_pbrsb rfds bhi
bogomips : 5836.80
clflush size : 64
cache_alignment : 64
address sizes : 46 bits physical, 48 bits virtual```The one that takes just over an hour is running this CPU:
vendor_id : GenuineIntel
cpu family : 6
model : 154
model name : 12th Gen Intel(R) Core(TM) i7-1260P
stepping : 3
microcode : 0x421
cpu MHz : 2959.435
cache size : 18432 KB
physical id : 0
siblings : 16
core id : 23
cpu cores : 12
apicid : 46
initial apicid : 46
fpu : yes
fpu_exception : yes
cpuid level : 32
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect user_shstk avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req hfi vnmi umip pku ospke waitpkg gfni vaes vpclmulqdq rdpid movdiri movdir64b fsrm md_clear serialize arch_lbr ibt flush_l1d arch_capabilities
vmx flags : vnmi preemption_timer posted_intr invvpid ept_x_only ept_ad ept_1gb flexpriority apicv tsc_offset vtpr mtf vapic ept vpid unrestricted_guest vapic_reg vid ple shadow_vmcs ept_mode_based_exec tsc_scaling usr_wait_pause
bugs : spectre_v1 spectre_v2 spec_store_bypass swapgs eibrs_pbrsb rfds bhi
bogomips : 4992.00
clflush size : 64
cache_alignment : 64
address sizes : 39 bits physical, 48 bits virtual
power management:```The i9 is taking WAY longer. I'm not great with hardware specs, but I really would have thought the newer i9 would do better than the i7. Clearly not though.
barren mist
So it's not that the same config in the same environment used to take less time than it does now, it's that the same (?) config in two completely different environments show different runtimes?
barren mist
Right, but it's not the same environment that "used to" take an hour and now takes 5.
It's one environment that has always taken an hour (and still does?) vs a different environment that has always (?) taken five hours.
worldly cradle
No, apparently the VM migrated through my system to a different node which I didn't notice until I went digging. That's my bad. That said, a 4x difference on a "better" CPU is somewhat peculiar, no?
barren mist
No idea without any details on the rest of the environment. Maybe disk access on the new one is way slower and creating all those temp overlay images is taking four times longer.
worldly cradle
Both of them are connected via 10Gbit connection to my NAS, via iSCSI.
barren mist
If these are two completely different pieces of hardware it seems there are lots of opportunities for differences between the environments.
Seems unlikely to be a Kometa code or config issue.
I suppose you could run some trace logs to see if that provides any clues and maybe recreate the old environment to see if it still performs the same, then start the detailed comparisons.
worldly cradle
I did run the old environment to test, and it was 6+ hours.
barren mist
Is that last log above one of the old one-hour runs?
worldly cradle
Yes.
barren mist
FWIW, sample size of one, same config [yours without the external YAML files] against the same plex on the same machine under 8 Kometa versions and 10 docker images from over the last year:
kometa:v1.21.1 [INFO] | Finished: 05:14:58 2025-06-08 Run Time: 0:39:29 |
kometa:v2.0.0 [INFO] | Finished: 04:26:47 2025-06-08 Run Time: 0:45:58 |
kometa:v2.0.1 [INFO] | Finished: 03:31:31 2025-06-08 Run Time: 0:51:53 |
kometa:v2.0.2 [INFO] | Finished: 02:29:08 2025-06-08 Run Time: 0:54:34 |
kometa:v2.1.0 [INFO] | Finished: 01:25:03 2025-06-08 Run Time: 1:06:17 |
kometa:v2.2.0 [INFO] | Start Time: 23:17:44 2025-06-07 Finished: 00:07:16 2025-06-08 Run Time: 0:49:31 |
kometa:develop [INFO] | Start Time: 19:26:06 2025-06-07 Finished: 20:12:18 2025-06-07 Run Time: 0:46:11 |
kometa:latest [INFO] | Start Time: 22:22:08 2025-06-07 Finished: 23:08:26 2025-06-07 Run Time: 0:46:17 |
kometa [INFO] | Start Time: 16:51:21 2025-06-07 Finished: 17:39:34 2025-06-07 Run Time: 0:48:13 |
kometa:nightly [INFO] | Start Time: 20:18:17 2025-06-07 Finished: 20:59:18 2025-06-07 Run Time: 0:41:01 |
Turns out not the problem that is actually being chased here, but here's the data.
barren mist
No idea why specifically, but everything is taking longer; TV overlays are the main issue though
1 hour run
resolution 00:04:xx
status 00:02:3x
audio 00:09:38
ratings 00:02:50
video 00:05:24
runtime 00:02:2x
checking for overlaid posters 00:00:25
overlay application prep 00:04:00
overlay application 00:18:xx
6 hour run
resolution/edition 00:26:xx
status 00:05:xx
audio codec 01:04:xx
rating 00:18:xx
video codec 00:36:xx
runtime 00:17:xx
checking for overlaid posters 00:02:xx
overlay application prep 00:20:00
overlay application 01:45:xx
So do you think this may be a bug of some sort?
barren mist
It seems unlikely since I cannot reproduce this at all. All ten of my runs across the entire version history from the last year are within a few minutes of each other.
And there have been no code changes on the kometa side between your fast-then-slow runs.
If it's a bug, it's being triggered by something peculiar to your setup.
In this section here:
overlay application 00:18:xx
vs
overlay application 01:45:xx
Very few overlays are actually applied, so almost the only thing going on here in either case is communication with Plex and queries against Kometa's cache database.
The fast-vs-slow runs are against pretty much the same number of items, and similar numbers of overlays are actually applied.
worldly cradle
No idea why specifically, but *everything* is taking longer; TV overlays are the...
I misunderstood the context in which you were making this statement.
barren mist
Looking at one specific overlay:
video codec 00:05:24
vs
video codec 00:36:xx
It seems like communication with Plex is a big part of the issue:
1 hour run
0 second search for all shows
20 seconds filtering for UHD
20 seconds to load all episodes
4 seconds filtering for REMUX
20 seconds to load all episodes
15 seconds filtering for BLU-RAY
20 seconds to load all episodes
1 min 5 seconds filtering for WEB
20 seconds to load all episodes
20 seconds filtering for HDTV
20 seconds to load all episodes
12 seconds filtering for DVD
20 seconds to load all episodes
10 seconds filtering for SDTV
20 seconds to load all episodes
1 second filtering for TELESYNC
20 seconds to load all episodes
1 second filtering for CAM
6 hour run:
0 second search for all shows
2 minute filter for UHD
2 minutes loading all episodes
30 seconds filtering for REMUX
2 minutes loading all episodes
2 minutes filtering for BLU-RAY
2 minutes loading all episodes
8 minutes filtering for WEB
2 minutes loading all episodes
2 minutes filtering for HDTV
2 minutes loading all episodes
1.5 minutes filtering for DVD
2 minutes loading all episodes
1.5 minutes filtering for SDTV
2 minutes loading all episodes
10 sec filtering for TELESYNC
2 minutes loading all episodes
10 sec filtering for CAM
So perhaps something has changed there?
split cloudBOT
antwanchild used !afnh
@worldly cradle, anything further needed here? If not, please type /close and hit enter. Please respond within 24 hours of this message or it will be archived.