We are doing tests on invoicing. We have 219 validated invoices in February. We have activated the trytond-worker.
If we select the whole 219 invoices and we do a post on it, it’s very slow. (sometimes we have a warning : “The invoice … generates a payment date … in the past”) Between 2 warnings, there’s approximately 1 minute ! (and warning concerns 2 consecutive invoices (we check invoice numbers displayed in the warning).
If we select a block of 20 invoices, time to post is approximately 1 minute (which is correct).
So, wondering why it takes so much time when selecting all the invoices ? (something wrong with the process to check numbering?) Problem with the worker ?
I think its relevant to share the number of lines that it has each invoice. From my experience its not the same to post 219 invoices with 1 line, than posting 219 invoices with 30-40 lines each. The latter will be slower.
Could you share such detail so others can test also?
When a warning is raised, the all request is roll-backed so if you skip the warning the server has to restart the work from beginning.
This means that it takes 1 minute to reach the warning which seems strange because the only operations before that are setting employee (a single write) and grouping by company (a single read).
So I would suspect that the validation after the write of employee is slow.
It will be interesting to see where time is spent.