Results returned have dropped from 22K to 17K per day. and those would have been done by fast machines so more priority units for us ordinary crunchers.

Mike.

I'm skeptical of the described problem in 4572. I currently run 64 concurrent ARP1 work units and have done so for the past 28 days (broke away from ARP1 to run MCM for a challenge). Have also run 128 concurrent (cut back due to the length of time to complete a WU). I have never encountered the issue described after contributing over 100 years of computer time to ARP1. I have seen the uploads fail and accumulate like during system maintenance windows or when WCG was encountering filesystem errors on the upload storage device (resulting is a lot of HTTP errors). During those times, I did see the same messages as described in the incident but was able to retry the uploads and get them to clear. Yes, if enough accumulated in upload pending status the downloads would cease but at no time did the BOINCMGR disconnect or require a reboot or client restart to clear. In the past there have been very rare instances where, due to circumstances, the upload process got interrupted and would not restart without intervention by the WCG staff (like to delete the upload file from the upload filesystem so that the client and remote end were back in sync) and the upload would finish as normal. I have only seen this documented 4 or 5 times in 14 years was usually due to a power outage or similar immediate disconnect that took down not only the client but also the OS. I have also encountered times where I have lost internet connectivity and the uploads accumulated on the client until a connection was reestablished (maybe 24 hours or more later). Once the connection was established a flood of large file uploads would commence (I have mine set to 10 uploads concurrent) and would complete without a problem. Yes, it took as long as an hour sometimes but they did complete without intervention.

Kevin's last report said that there were 130 units currently stuck, having errored out. 96 of these would seem to be what remains of generations 079 - 095 plus most of 098 & 099.

I will look a little deeper into this for my next report at the weekend.

Mike

My apologies for the distraction. I ran into connectivity issues with my 16 thread system and a similar logjam. I found that report by chance and since the reporter had expressed an interest to push ARP at a rate of 2000 WUs/day, I thought they (and other similar contributors) may have work sitting there attempting to be sent.

I've bumped my ARP limits from 3 to 6 on the main cruncher (16 threads/15 active) and from 1 to 2 on the laptops (4 threads each). So far, there doesn't seem to be any issues, but it will mean more lost work when I need to reboot them.