Shows picture.

When you look at the above picture, you'll see I have 1 Enigma task at this moment, but two are in memory.
Earlier today I had 12 Enigma applications in memory, so it would seem it doesn't unload itself when the task is done.

This is on a Windows 2000 SP4, BOINC 5.10.26 system. Earlier was 5.10.24 (alpha testing).

---

Jord.

BOINC FAQ Service.

Oh, it's worse than that. It even stays in memory when I exit BOINC completely. Trying to end the application in Task Manager gives me an error. Access is denied.

---

Jord.

BOINC FAQ Service.

That's weird, I haven't seen any problems on my Win2k test machine.
Have you checked who's the owner of the second process (maybe someone else installed BOINC on that host :D ) ? Does it take any CPU time ?

---

M4 Project homepage
M4 Project wiki

None of them take any CPU time and owner is me.
It's just gotten worse though and I posted it to the BOINC Alpha email list as well, as now I had 15 Enigma's, 2 Malaria's and 3 Cosmology apps in memory, all not taking CPU cycles and all staying when I exited BOINC. A reboot helped there.

I'm suspecting it is the wrapper that's doing this.

---

Jord.

BOINC FAQ Service.

My wrapper was build from svn revision 13449, not modified. I used Visual Studio 2005 express as a compiler.

---

M4 Project homepage
M4 Project wiki

I'll keep an eye on it. It's weird though that the wrapper isn't staying in memory, only the applications are. And this only started happening since BOINC 5.10.24, so it's got to be something between that and the wrapper, as I didn't have the problem before.

But just look at the latest picture. Any more and I get BOINC complaining it doesn't have memory to run apps in. ;)

---

Jord.

BOINC FAQ Service.

Hi,
can you try it also with yoyo@home and hydrogen@home? Both projects using also the wrapper.
yoyo

---

Hi,
can you try it also with yoyo@home and hydrogen@home? Both projects using also the wrapper.
yoyo

I'm already testing it with H2. Won't do yours again for the moment as I had too many (stability) problems (with my PC) with your tasks. Sorry.

The H2 tasks run in one go and without a problem. They unload correctly. For now running with the <task_debug/> flag on. I had the feeling it was the TSP project that was causing this, but last night I ran without enabling it asking for work and that ended with those 15 Enigma's etc. in memory this morning. So I'm a bit lost. Now doing an hour run with TSP enabled.

---

Jord.

BOINC FAQ Service.

OK, interesting.

I have 1 Enigma task at this moment. Yet it's in memory twice!
See this picture. It ran for 1m 44s first, then was suspended because another TSP task with a closer deadline needed to be done first. That went on for a while, all kinds of apps switching in and out of memory. Then Enigma restarted, but instead of using the application that was in memory already, it started a new one, which ran for 11 seconds.

It now shows in BOINC Manager to have run for 11 seconds, not 1m 55s !

---

Jord.

BOINC FAQ Service.

I now forced all other tasks to suspend and only let Enigma run. It now continues off from the second entry that it had, the one on 11 seconds.

**confused**

---

Jord.

BOINC FAQ Service.

* TJM is downloading 5.10.26

Ageless wrote:

It now shows in BOINC Manager to have run for 11 seconds, not 1m 55s !

That's nothing unusual, for some reason CPU time is not saved when the task is stopped.

---

M4 Project homepage
M4 Project wiki

Maybe it's a conflict between my wrapper and the other two in your memory :D

---

M4 Project homepage
M4 Project wiki

Maybe it's a conflict between my wrapper and the other two in your memory :D

That would count for Enigma staying in memory, but I doubt it'll wreck the capability of two other projects to unload in a normal way.

It's probably two bugs, though.
One is where Enigma doesn't know it has started already once before and just starts a new application, instead of resuming the old one. Probably to do with it checkpointing internally, not outside the wrapper. (??)

Two is where applications just do not unload from memory when they are completely through. The wrapper does unload correctly at all times.

I'm also posting all my results and findings on this to David Anderson.

---

Jord.

BOINC FAQ Service.

Ageless wrote:

That would count for Enigma staying in memory, but I doubt it'll wreck the capability of two other projects to unload in a normal way.

It's probably two bugs, though.
One is where Enigma doesn't know it has started already once before and just starts a new application, instead of resuming the old one. Probably to do with it checkpointing internally, not outside the wrapper. (??)

The main app does not start/stop by itself. The wrapper does that, it starts a fresh enigma with a command line -R -o results.txt 00trigr.naval 00bigr.naval or resumes one which is already in memory.

Ageless wrote:

Two is where applications just do not unload from memory when they are completely through. The wrapper does unload correctly at all times.

Or the wrapper is dying for some reason, leaving suspended app in memory...

< Last Edit: October 23, 2007, 02:29:45 pm by TJM >

---

M4 Project homepage
M4 Project wiki

TJM wrote:

Or the wrapper is dying for some reason, leaving suspended app in memory...

Heh, that one is for you to hunt down and fix, then.

Any other cc_config.xml flags you want me to turn on?

---

Jord.

BOINC FAQ Service.

So far we've been unable to reproduce the bug (but we are still working on it), it *might* be specific to your system.

---

Thanks. I always were a bit strange. :P

I see it happen on long runs, not noticeably when I sit here with my nose on top of things.
Got the extra flags on, that's going to be fun on my log files. I also suspended all the other projects and let only Enigma run, with TSP trying to get work. Let's see if I can force it in this way.

I'll report back in an hour if I have anything to report and else I'll continue all other projects as well and let it rip through the night. I had the 'problem' two days in a row, it would be strange if it didn't happen the third night.

Edit: Let's just say I'll let it run through the night, unattended as I just see TSP is down and thus not giving out work. I don't know of another project this quickly that has such fast tasks. :(

---

Jord.

BOINC FAQ Service.

This reminds me a weird bug I've seen in 5.10.13 - after about 20 days uptime my slowest PC suddenly started acting like a dual cpu machine. It ran two tasks simultaneously, both were consuming CPU time (about 50% each), but tasks runtimes were 10-15 times longer than usual (average ~1h40m task took 20+ hours to complete). I had to restart the system, because after I shut down manager I couldn't start it again.

---

M4 Project homepage
M4 Project wiki

Ageless wrote:

Edit: Let's just say I'll let it run through the night, unattended as I just see TSP is down and thus not giving out work. I don't know of another project this quickly that has such fast tasks. :(

I have TSP attached, and I've just received some work units. Btw, one task is stuck (100% done) with a 'Waiting for memory' message - it's the first time I see such a message, what does that mean ?

---

M4 Project homepage
M4 Project wiki