You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
From the conversation on discord (https://discord.com/channels/781145214752129095/1081743238194536458/1291430902450356285):
I noticed something while trying to add this to my worker. It seems that only requests/images that are in "waiting" can be canceled, so if any one of those goes from processing to restarted and back to processing endlessly then it can not be canceled. This can happen if the request has "is_possible": true
Managed to make my worker pseudo-crash, it didn't stop the execution like I wanted, but the processes crashed
The problem was that on restarting it, it would go on to pick that request again, even though I tried VERY hard to cancel it in https://aihorde.net/api/ with the delete endpoint /v2/generate/status/{id}, but it didn't work, the request could only be canceled when I stopped my worker completely and "is_possible" changed to false
quoting db0:
"Actually, any request will be cancelled. Cancelling a request sets its n to 0. However you're right that failing a cancelled request is an edgecase I hadn't considered as a reported fault on the horde does an n += 1. But theoretically should abort after 3 retries. Need to double-check that logic. If you can open an issue I'll follow-up"
The text was updated successfully, but these errors were encountered:
db0
added a commit
that referenced
this issue
Oct 5, 2024
From the conversation on discord (https://discord.com/channels/781145214752129095/1081743238194536458/1291430902450356285):
I noticed something while trying to add this to my worker. It seems that only requests/images that are in "waiting" can be canceled, so if any one of those goes from processing to restarted and back to processing endlessly then it can not be canceled. This can happen if the request has
"is_possible": true
Managed to make my worker pseudo-crash, it didn't stop the execution like I wanted, but the processes crashed
The problem was that on restarting it, it would go on to pick that request again, even though I tried VERY hard to cancel it in https://aihorde.net/api/ with the delete endpoint
/v2/generate/status/{id}
, but it didn't work, the request could only be canceled when I stopped my worker completely and"is_possible"
changed to falsequoting db0:
"Actually, any request will be cancelled. Cancelling a request sets its n to 0. However you're right that failing a cancelled request is an edgecase I hadn't considered as a reported fault on the horde does an
n += 1
. But theoretically should abort after 3 retries. Need to double-check that logic. If you can open an issue I'll follow-up"The text was updated successfully, but these errors were encountered: