Getting wrong WUs
31 Dec 2021 23:21:56 UTC
Topic 226684
(moderation:
Hi,
I disabled the Gravitational Wave search applications and set 'Allow non-preferred apps' to 'No', but I still always get Gravitational Wave search applications. What's going on here?
Cheers
Hype
Are you matching your changed
)
Are you matching your changed preferences to the host location or venue? Make sure your host is assigned to the same venue that you have changed your preferences in. There are four possible host locations in the Settings.
Restart BOINC to make sure you pick up the new location and changed preferences.
Hype wrote: Hi, I disabled
)
Make sure you have unchecked everything in your preferences except Gamma-Ray GPU tasks.
Also, in your Boinc Manager, do an update after unchecking (above) your preferences.
If that does not help, check your config files and your local preferences files and make sure you don't have any Gravitational Wave item selected in your files.
Also, do as Keith suggested, too.
Proud member of the Old Farts Association
Thanks, I tried your
)
Thanks, I tried your suggestions but it doesn't work.
My 'Generic' preferences:
'Generic' profile on the machine:
Restarted computer to make sure and instantly got 24 Grav search WUs:
I have the GPUs running in a separate instance, but they're merged into the 'AQUA' host on Einstein.
What do you mean by local preferences files? app_config? I don't have any for Einstein.
So, it seems to be a problem
)
So, it seems to be a problem related to my second instance.
On the first instance it works fine. Although the instances are merged into one on Einstein?
Hype wrote:So, it seems to
)
Between your two computers, I do see that you are receiving Gamma-Ray tasks. The first one, computer #12913324 has one task from last night, near midnight. The rest of them were "sent" to your computer approximately 10 hr earlier. The second one, computer #12913405 seems to be doing just fine.
You may be experiencing a delay in seeing the tasks because of the server's lag time in posting them to your computers. Give them a day or two and we will and see.
Proud member of the Old Farts Association
Hype wrote: What do you mean
)
In Windows 10, go to the drive that has BOINC installed on it. Looking at the folders on that drive, make sure you have the hidden files exposed. Click on ProgramData, then BOINC, and it should open to three folders, 'notices', projects', and 'slots'.
Below those three folders you will find "cc_config.xml", open it with notepad, but be careful not to make any changes. Just read the file and become familiar with it.
Next, click on the folder 'projects'. Then click on the 'einstein.phys.uwm.edu' folder. You SHOULD then find your 'app_config.xml' file. If you don't see it, then study these:
https://boinc.berkeley.edu/wiki/User_manual
https://boinc.berkeley.edu/wiki/Client_configuration
If you have anymore questions, continue asking. That's what we're here for.
Proud member of the Old Farts Association
If you want to make an
)
If you want to make an 'app_config.xml' file for Einstein>Projects, try this.
<app_config>
<app>
<name>hsgamma_FGRPB1G</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
<max_concurrent>2</max_concurrent>
</app>
</app_config>
Since you already are using 2 x 1080Ti GPUs, make sure you have '<max_concurrent>2</max_concurrent>' in the file. Also, make sure you only have one 'hsgamma_FGRPB1G' task running per GPU and CPU. It slows down the task time if you have more than one.
After doing all of that, make sure you UPDATE Einstein in BOINC Manager.
Proud member of the Old Farts Association
Hype wrote:I disabled
)
The crucial bit of information that you don't mention is what you did with the already downloaded tasks that you had at the time you changed from the GW search to the GRP search.
What you are supposed to do is abort and report any existing GW tasks so that the scheduler knows you no longer wish to crunch them. What I'm guessing is that you didn't do that but simply got rid of those tasks by perhaps resetting the project using that control in BOINC Manager.
If you look at your hosts list on the website, there is a link to the last scheduler contact where you can see the details of the most recent exchange between a particular host and the project servers. I was fortunate enough to pull up a page with entries dated 2021-12-31 23:19:40 UTC which (with a lot of other stuff) showed 12 entries just like this example.
[HOST#12913324] MSG(high) Resent lost task h1_0636.80_O3aC01Cl1In0__O3AS1a_637.00Hz_52300_2This supports the guess that you didn't abort, but simply used the 'reset' function (or some other method) to get rid of unwanted tasks.
If you reset the project, your client throws away all existing programs, data files, tasks, etc and downloads fresh copies of everything. It's quite a drastic action designed to give a fresh start when things have become really screwed up. It's quite unnecessary for just a simple preference change from one search type to another.
As you download all the new copies of everything, the scheduler will kindly re-send all the details of the existing tasks that you had on board in batches of 12 at a time with each scheduler contact. That is precisely why you must abort unwanted tasks before you take actions like a project reset. The only way to get the scheduler to "forget" the previously allocated tasks is to abort them.
Setting 'No new tasks' (and a small initial work cache size) is also advisable until you are sure that you are getting the correct tasks as expected for the new search.
Cheers,
Gary.
I checked it again
)
I checked it again today.
I always abort all the tasks, I didn't reset a BOINC instance before aborting the tasks.
The scheduler always resends these lost tasks, I can see that in the log.
Even if I abort them, update and wait. It will send them again.
Hype wrote:I always abort all
)
If you look at what I wrote, I specifically mentioned abort AND report so that the scheduler knows and can take action to send those tasks to someone else. If I need to do that, I always make sure the aborted tasks have been reported so there's no chance of the scheduler not knowing their true status.
When you abort tasks, only your client knows that fact. It doesn't necessarily get auto-reported immediately. It could take some time so it's important to force an update. I don't know from actual experience how the scheduler treats the project reset function. In all my years of running BOINC, I've never seen the need to reset the project so have never directly observed what happens.
My guess is that the scheduler, on receipt of a project reset, just assumes that it should send everything you previously had (according to the scheduler's own records) irrespective of the local status at your client end. I guess that a reset is treated as if everything at the client end may be screwed up and not necessarily trustworthy. So if you were to abort and immediately use reset without updating and waiting for confirmation, then I imagine what you describe might happen.
In many years of using the 'resend lost tasks' feature in a variety of ways, I've never, ever seen an example of a properly aborted and reported task being received again as a 'lost task'.
In that case I must assume that, for whatever reason, the scheduler wasn't told that those tasks were aborted. The scheduler simply doesn't resend tasks that have been reported to it as aborted.
Cheers,
Gary.