Clean Sweep of Team Firsts?

[Al]

Bryan got all the user firsts and the team firsts have been a sweep for SUSA thus far. The only one left is 1st team to 1 billion. As of today we are well in the lead to that goal and are 300 million away on Goofy CPU. Let's finish it off! Bryan, Mumps and EG have done their part with 250, 100 and 100 million respectively. I was planning to stop at 50 million in a few days, but I'll stay on it for another 50 million if a few more of us commit to taking our milestone to the next level or two or three. Anyone else willing to commit?

Edit: I'll do 50 million more, so we need an addition 250 million.

[Al]

Is anyone else seeing this?

Stderr output
<core_client_version>7.6.33</core_client_version>
<![CDATA[
<message>
finish file present too long
</message>
<stderr_txt>
13:07:53 (28076): wrapper (7.7.26016): starting
13:07:53 (28076): wrapper: running htest_64bits.exe (input)

================================================== =====
|HTEST v2.0.0 [22-02-2018] |
|Written by Lukasz Swierczewski |
================================================== =====


INFO: You do have the Intel CPU but
does not support the required instructions.
You can not use a hardware random number generator.
System use software generator.
13:13:46 (28076): htest_64bits.exe exited; CPU time 345.667016
13:13:46 (28076): called boinc_finish(0)

</stderr_txt>
]]>

This is my only machine with any errors, but it has completed Valid (10777) tasks. I'm only seeing this on one T5500 running windows.

[Ron Shurtz]

Throwing in my drop in the bucket. (20 threads)

[scole of TSBT]

Is anyone else seeing this?

Stderr output
<core_client_version>7.6.33</core_client_version>
<=!=[=C=D=A=T=A=[
<message>
finish file present too long
</message>
<stderr_txt>
13:07:53 (28076): wrapper (7.7.26016): starting
13:07:53 (28076): wrapper: running htest_64bits.exe (input)

================================================== =====
|HTEST v2.0.0 [22-02-2018] |
|Written by Lukasz Swierczewski |
================================================== =====


INFO: You do have the Intel CPU but
does not support the required instructions.
You can not use a hardware random number generator.
System use software generator.
13:13:46 (28076): htest_64bits.exe exited; CPU time 345.667016
13:13:46 (28076): called boinc_finish(0)

</stderr_txt>
]=]=>

This is my only machine with any errors, but it has completed Valid (10777) tasks. I'm only seeing this on one T5500 running windows.

If the issue still exists, this project has a lot of I/O and causes issues. I'm still convinced that a single boinc client can only handle so much I/O overhead. Try cutting the threads back to 23 or 22, leaving 1-2 free. On my high thread systems, I used to have to run 4 clients with at least a couple free on the machine.

[Al]

If the issue still exists, this project has a lot of I/O and causes issues. I'm still convinced that a single boinc client can only handle so much I/O overhead. Try cutting the threads back to 23 or 22, leaving 1-2 free. On my high thread systems, I used to have to run 4 clients with at least a couple free on the machine.

I'll give it a shot.

Sent from my VS501 using Tapatalk

[Bryan]

"The Finish File Present Too Long" can be caused by too many WU finishing at exactly the same time. A single BOINC client doesn't have enough time to do all the clean up - stop the WU, generate the output file, close the slot, and erase the WU from the HDD. It doesn't happen that often on the 24 thread machine but it DOES happen if too many WU complete at the same time.

There are 2 fixes:
1. Run 2 BOINC clients.
2. SUSPEND all but 6-8 WU. After one minute RESUME 6-8 more WU. One minute later RESUME the rest.

This is a VERY common problem on the higher thread count machines.

[Al]

"The Finish File Present Too Long" can be caused by too many WU finishing at exactly the same time. A single BOINC client doesn't have enough time to do all the clean up - stop the WU, generate the output file, close the slot, and erase the WU from the HDD. It doesn't happen that often on the 24 thread machine but it DOES happen if too many WU complete at the same time.

There are 2 fixes:
1. Run 2 BOINC clients.
2. SUSPEND all but 6-8 WU. After one minute RESUME 6-8 more WU. One minute later RESUME the rest.

This is a VERY common problem on the higher thread count machines.

I've been running it for several days now and the wus are spread, 3 or 4 finishing at any one time. I'm thinking this may not be my problem, though I'll keep an eye on it. I did drop down to 23 threads and have had no more computational errors. I'll see if that holds up overnight.

[Al]

Throwing in my drop in the bucket. (20 threads)

Thanks Ron, what milestone will you go after? Looks like 25 million is your next one...7 million will get you there.

[Al]

Dropping active threads to 23 decreased the amount of errors, but didn't eliminate them. I've lowered the active threads to 22 to see if that will do the trick.

[FourOh]

A challenge where I don't have to make any effort to participate? Perfect! I'm pretty much on cruise control these days due to RL stuff, so I'll stay on it for a while.