Re: WU 11758 & 11778 [Fixed]
Posted: Sun Mar 29, 2020 3:09 pm
Yes, Niel-B is correct.
All WUs must be sent back to the Work Server that is running that particular project. The Collection Server is just a cache for WUs which cannot reach the WS in a timely manner. (It's the back-end server).
Assuming the WU is successfully uploaded to a CS, the WU receives an official time of completion stamp. Then when the CS can establish a connection to the WS, the WUs which have accumulated on CS are forwarded to the WS and the credit record is sent to the DB server.
Due to the oveerwhelming load being placed on our servers' banwidth, the CS is having difficulties forwarding it's cache so it continues to grow and the eventual arrival of the credit record is waiting even longer, making scores of Donors unhappy. The stats DB, is also congested, which doesn't help but that's what people notice.
With considerable effort, we have up-scaled the distribution of work units to a phenomenal level (and people are still working on that) but that has put a lot of pressure on other components of the back-end servers and their processes.
All WUs must be sent back to the Work Server that is running that particular project. The Collection Server is just a cache for WUs which cannot reach the WS in a timely manner. (It's the back-end server).
Assuming the WU is successfully uploaded to a CS, the WU receives an official time of completion stamp. Then when the CS can establish a connection to the WS, the WUs which have accumulated on CS are forwarded to the WS and the credit record is sent to the DB server.
Due to the oveerwhelming load being placed on our servers' banwidth, the CS is having difficulties forwarding it's cache so it continues to grow and the eventual arrival of the credit record is waiting even longer, making scores of Donors unhappy. The stats DB, is also congested, which doesn't help but that's what people notice.
With considerable effort, we have up-scaled the distribution of work units to a phenomenal level (and people are still working on that) but that has put a lot of pressure on other components of the back-end servers and their processes.