"Could not load model 143" warning flooding the logs about suspicious_login HOT 8 CLOSED

BTW, and sorry if it is a silly question but, if only the last model is used (in my case 146 I guess), are the previous models removed at some point or are they used to get the new models? Will models accumulate indefinitely?

Known unresolved enhancement: #51

from suspicious_login.

same issue here, with NC 20.0.1

did the upgrade yesterday from NC 19.0.6 .. got Internal Server Error and could not login again.
Then i had to rebuild the "Instanceid" .. when doing that, the appdata_xxxx Folder will also being rebuilt.

Since then there is the following error:

[suspicious_login] Fatal: Could not load classifier model 561: /appdata_ocxxxxx/suspicious_login/models

REPORT /remote.php/dav/addressbooks/users/xxxxx/contacts/
from 1xxxxxxx by xxxxx at 2021-04-15T11:07:42+02:00

Since then i have got this error

running command:

occ suspiciouslogin:predict xxxx xxx.xxx.xxx.xxx

In ModelStore.php line 135:

Could not load model 561

In Root.php line 205:

/appdata_ocxxxxxxxx/suspicious_login/models

from suspicious_login.

You should be able to resolve this with occ suspiciouslogin:train

from suspicious_login.

think there is something that can be improved here.

from suspicious_login.

You should be able to resolve this with occ suspiciouslogin:train

Unfortunately the train didn't seem to help, at least when using predict from the CLI is still throwing the error:

$ occ suspiciouslogin:train
Using ipv4 strategy
Prescision(y): 0.99019607843137
Prescision(n): 1
Recall(y): 1
Recall(n): 0.99009900990099
$ ncc suspiciouslogin:predict luca 192.12.12.12 143

In ModelPersistenceService.php line 138:
                            
  Could not load model 143  
                            

In Root.php line 205:
                                                     
  /appdata_xxxxx/suspicious_login/models/143  
                                                     

suspiciouslogin:predict [--v6] [--] <uid> <ip> [<model>]

But I guess the error handling is missing and a model that can't be saved to storage should be cleared from the database.

I see, so it is a possible scenario that the out of space error leaded to this situation. I hope some error handling can be added sooner than later :)

from suspicious_login.

Try ncc suspiciouslogin:predict luca 192.12.12.12. Model 143 doesn't work but the app will always take the latest model available anyway.

from suspicious_login.

OK, that didn't show any errors, but I'm not sure if it would have shown any errors before the train either :)

I checked the DB and model 143 is still there, and there is no 143 file in appdata_xxxxx/suspicious_login/models/143 (but there is a new file 146. I will keep monitoring the logs to see if the error is really gone.

BTW, and sorry if it is a silly question but, if only the last model is used (in my case 146 I guess), are the previous models removed at some point or are they used to get the new models? Will models accumulate indefinitely?

Thanks!

from suspicious_login.

I checked the DB and model 143 is still there, and there is no 143 file in appdata_xxxxx/suspicious_login/models/143 (but there is a new file 146. I will keep monitoring the logs to see if the error is really gone.

@Githopp192 is that also the case for you – the file doesn't exist, right?

from suspicious_login.