too many open files ? #983
Comments
|
Are there any zombie processes? If no, check the current ulimit and play around with it. |
ilike2burnthing
added
the
more information needed
|
Hello, I have the same problem. After a few days I see this in the logs As soon as this happens all requests will fail until I reboot the container. I do not see any zombie processes |
|
|
I'm also having this problem. |
|
I updated my open files limit with Will give feedback in a few days if the problem happens again. |
|
I still the same problem. Error: Error solving the challenge. [Errno 24] Too many open files no zombie processes ps -aux | grep Z
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
root 650961 0.0 0.1 3324 1536 pts/1 S+ 16:34 0:00 grep Z
root@prowlarr:~# no limits: ulimit cat /proc/sys/fs/file-max
9223372036854775807
root@prowlarr:~# ulimit -Hn
1048576
root@prowlarr:~# ulimit -Sn
1024
root@prowlarr:~# cat /proc/sys/fs/file-max
9223372036854775807Only way is to restart the service I created with : any idea how I can get more info to debug ? |
ilike2burnthing
added
help wanted
|
See if v3.3.11 makes a difference. |
|
to fix just add in file This is because Python stores the process metadata even after it terminates |
|
If those above can test and confirm this is working, I'll add it in the next release. |
|
Which version can we install to test this fix ? |
|
Run from source - https://github.com/FlareSolverr/FlareSolverr#from-source-code Or if using Docker, edit the file in the container and restart. |
ilike2burnthing
added
the
more information needed
|
For now I have installed 3.3.12 because I'm in a lightweight headless LXC container running flaresolverr binary distribution and I don't want to pollute this container by installing chrome and all it's dependencies just to run the source version of flaresolverr. I might try in the future in another container. Anyway if you think the process should be closed at that point why not doing process.close() ? Maybe you have a reason, I did not look at the source code yet ? |
|
As I can't reproduce the issue I can't test if this resolves it. |
|
I adapted the code as suggested. As this problem only occurs after a few days I'll give you a feedback when relevant (did the change yesterday). |
|
Easy way to reproduce - use windows version, because of 512 MOF in windows, just send 515 requests. |
It's original undetected_cromedriver package, which rarely updates |
There are a few modifications, but yes, 99.9% of UC is untouched. |
I also change the code as suggested déc 22 on version 3.3.12 normaly the error occur 4/5 days after container start and never reach 1 week of uptime so i will update you next week if it solve |
|
Got TMOF, but time to error enlarged from 1 day to 2 days with ~ same amount of requests per day |
|
After 6 days, not a single failure :). It usually started failing after 2, 3 days for me. |
|
Hi the team, do you know when the app will be updated? I have a Docker version and I don't quite understand how to change the parameter to overcome this problem ^^' thks U |
|
@LaCartouche how are your instance of Flaresolver working? Did you have any issues? |
|
I just took a second look at my logs. It is all working BUT I think Yggtorrent disabled their captcha verification. From the logs: 2023-12-27 12:34:53 INFO Challenge not detected! If there is no captcha to solve I don't know if the undetected_chromedriver is used or not (I guess not ?). Which would then explain why I have no issues anymore and invalidate my testing. |
|
Any time the container is redeployed, this would need to be repeated. |
|
I still have the "too many open files" issue even with the suggested fix. |
|
Here's the log of my last test with a container that was just restarted and with the suggested fix applied. 335 requests until the "OSError: [Errno 24] Too many open files" |
|
9 days and still up |
|
iDope, TorrentQQ, and Torrent[CORE] are all public indexers with Cloudflare challenges, you could add them, configure them with Sonarr/Radarr/whatever you're using, see if the issue continues. |
|
Added one with a challenge i will keep you updated |
|
ok Now 6 Days with this fix and still up with challenge solved every 30 min |
|
ok issue happend today again but after 7 days compared to the usual 4 /5days so around 50% improvement Process Process-336: |
|
ok i have run lsof in the container and im flooded with this for maybe a thousand line python 7 flaresolverr 11u IPv4 4261517 0t0 TCP flaresolverr:8191->10.10.7.4:40246 (CLOSE_WAIT) |
|
Additional fix in file p = Popen([executable, *args], stdin=PIPE, stdout=PIPE, stderr=PIPE, **kwargs)to p = Popen([executable, *args], stdin=None, stdout=None, stderr=None, **kwargs)this have no impact on workflow, but reduce pipes usage, which usage lead to TMOF |
|
You know the drill people, test and let me know ;) |
|
Tried this new suggestion and started 2000 calls. I still have the same problem. Started from a fresh container, I had around 50 open files at startup. After running my test I'm stuck at 5197 open files.
|
|
My slow brain realized I had to switch to the python user... It's actually a variation of this line: python 7 169 python flaresolverr 877u IPv4 4359744 0t0 TCP localhost:46236->localhost:54065 (CLOSE_WAIT) |
|
May have something to do with the usage of kill while multiprocessing was starter so they become orphan |
|
ok found that the owner of all those file seam to be flaresolverr.py and this is the process 7 hope it help to narrow the issue |
|
ok cut 1/3 of those file got rid of those : python 7 flaresolverr 13u IPv4 7465067 0t0 TCP localhost:42260->localhost:48683 (CLOSE_WAIT) so now only have those i think its the fact to kill trhe detached browser from the outside that create this |
|
Please test proposed fix - ultrafunkamsterdam/undetected-chromedriver#1812. |
|
it work made the change run a bunch of solving test and I'm not spammed by open file Thanks |
|
If someone else can confirm this as well (@LaCartouche maybe?), then I'll add it. |
|
this is with the fix this is without fix after only few solve |
|
I can confirm after modifying the init.py file with the two lines suggested. I no longer have any "too many open files" errors. many thanks |
Have you checked our README?
Have you followed our Troubleshooting?
Is there already an issue for your problem?
Have you checked the discussions?
Environment
Description
after a few days of correct functionning, I get the error "internal server error" with "Too many open files"
Logged Error Messages
Screenshots
No response
The text was updated successfully, but these errors were encountered: