Only able to get 1 WU at a time now

Author	Message
RFGuy_KCCO Send message Joined: 26 Jun 20 Posts: 3 Credit: 100,433,137 RAC: 0	Message 131 - Posted: 1 Jul 2020, 20:49:11 UTC Last modified: 1 Jul 2020, 21:00:18 UTC All of my machines can only manage to download one task per machine beginning today. Is this by design, or is something amiss? Really hate having a GPU sit idle between the time one WU finishes and the next can be downloaded and started. ID: 131 ·

tito Send message Joined: 28 Jun 20 Posts: 5 Credit: 27,409,723 RAC: 1,647	Message 132 - Posted: 1 Jul 2020, 20:53:06 UTC Same here. ID: 132 ·

Bernie Vine Send message Joined: 26 Jun 20 Posts: 1 Credit: 2,425,603 RAC: 0	Message 144 - Posted: 2 Jul 2020, 4:31:10 UTC Yes here as well, but at least it has stopped the tasks running in priority mode because the deadline is so short. ID: 144 ·

chip Send message Joined: 14 Jun 20 Posts: 78 Credit: 1,321,619 RAC: 0	Message 150 - Posted: 2 Jul 2020, 13:28:42 UTC - in response to Message 131. Last modified: 2 Jul 2020, 13:31:23 UTC Frankly, I don’t fully understand why this is happening. The scheduler is set to 1 * NGPUs and it seems to be doing exactly that for several clients. Perhaps we just increase this value as a workaround, but in theory there shouldn’t be an issue as it is. ID: 150 ·

RFGuy_KCCO Send message Joined: 26 Jun 20 Posts: 3 Credit: 100,433,137 RAC: 0	Message 155 - Posted: 2 Jul 2020, 14:17:59 UTC All of my machines went back to getting 8 WU's at a time as of a short time ago. Thanks! ID: 155 ·

chip Send message Joined: 14 Jun 20 Posts: 78 Credit: 1,321,619 RAC: 0	Message 170 - Posted: 3 Jul 2020, 17:05:38 UTC - in response to Message 155. Good stuff, I'm going to lock this thread for now, because on review the value was actually set at 8* (nGPU + nCPU) which shouldn't only dish out one task. If this behaviour continues, please start a new thread. Thanks ID: 170 ·