i wish there were possible to suspend individual projects and not only entire boinc at once.

Yes i would also like a way to suspend or pause different projects. As for example the boinc beta is currently offline and boinc spends hours trying to connect to the project. And while its doing that every so often it decides to go on a "grab more work units" spree and even though i have 4 predictor work units cached decides to pull down another 4 work units. This is a bug in itself but if i had a way of disabling boinc beta (without detaching it) it would be helpfull.

Jamie

> Yes i would also like a way to suspend or pause different projects. As for
> example the boinc beta is currently offline and boinc spends hours trying to
> connect to the project. And while its doing that every so often it decides to
> go on a "grab more work units" spree and even though i have 4 predictor work
> units cached decides to pull down another 4 work units. This is a bug in
> itself but if i had a way of disabling boinc beta (without detaching it) it
> would be helpfull.

No, that's not a bug. It's the "low water mark" feature in the caching system. If you don't want it to do this, set your cache to something like 0.1-x days. With a low water mark of 0.1 days, It won't grab more WUs until you're empty. (Predictor might still grab some, as it's WUs are much shorter)

> No, that's not a bug. It's the "low water mark" feature in the caching system.
> If you don't want it to do this, set your cache to something like 0.1-x days.
> With a low water mark of 0.1 days, It won't grab more WUs until you're empty.
> (Predictor might still grab some, as it's WUs are much shorter)
>

Ok so if you call 8 work units a low water mark when its by far over the 0.2 days i have set in my prefs. I dont know what something like 0.02 days would give me. I have been a member of the boinc beta test for ages. I have been following the development of the application i dont need to be flamed for basic knowledge.

Now unlike most people here who are flaming berkeley for the last few days and im talking people who expect to install boinc and BOOM it works fine. Im sitting here patiently waiting for things to come into action. Im just making a suggestion that they could implement. Im not saying they have to im not saying they should and i also dont expect to get flamed for trying to say its something that i know it isnt.

I didnt mean that to sound nasty or offensive in anyway (if it has insulted anyone)

> Now unlike most people here who are flaming berkeley for the last few days and
> im talking people who expect to install boinc and BOOM it works fine. Im
> sitting here patiently waiting for things to come into action. Im just making
> a suggestion that they could implement. Im not saying they have to im not
> saying they should and i also dont expect to get flamed for trying to say its
> something that i know it isnt.
>
> I didnt mean that to sound nasty or offensive in anyway (if it has insulted
> anyone)

Errmmm.... How was my comment a flame? I sure hope I could do better than that! I was simply trying to help, forgive me.

There's a difference you know... If I wanted to flame you, I'd just call you a moron.

Here's a tip... Don't accuse someone of flaming then say you didn't mean to sound nasty or offensive.

> > Now unlike most people here who are flaming berkeley for the last few
> days and
> > im talking people who expect to install boinc and BOOM it works fine. Im
> > sitting here patiently waiting for things to come into action. Im just
> making
> > a suggestion that they could implement. Im not saying they have to im
> not
> > saying they should and i also dont expect to get flamed for trying to say
> its
> > something that i know it isnt.
> >
> > I didnt mean that to sound nasty or offensive in anyway (if it has
> insulted
> > anyone)
>
> Errmmm.... How was my comment a flame? I sure hope I could do better than
> that! I was simply trying to help, forgive me.
>
> There's a difference you know... If I wanted to flame you, I'd just call you a
> moron.
>
> Here's a tip... Don't accuse someone of flaming then say you didn't mean to
> sound nasty or offensive.
>

ok, fine im sorry, i guess i was a bit ott. Its what happens when you have just finnished reading a whole thread on people bitching at the seti@home team

I am in complete agreement with Darkspace and Jammie on this issue, though for somewhat different reasons. A new topic is open with a few suggestions of my own, including this one...

> > Yes i would also like a way to suspend or pause different projects. As
> for
> > example the boinc beta is currently offline and boinc spends hours trying
> to
> > connect to the project. And while its doing that every so often it
> decides to
> > go on a "grab more work units" spree and even though i have 4 predictor
> work
> > units cached decides to pull down another 4 work units. This is a bug in
> > itself but if i had a way of disabling boinc beta (without detaching it)
> it
> > would be helpfull.
>
> No, that's not a bug. It's the "low water mark" feature in the caching system.
> If you don't want it to do this, set your cache to something like 0.1-x days.
> With a low water mark of 0.1 days, It won't grab more WUs until you're empty.
> (Predictor might still grab some, as it's WUs are much shorter)
>
Actually, there is a bug. Sometimes when requesting work from a project that is disabled, BOINC then immediately requests 1 second of work from another project. Even if the queue is overfilled (above the highwater mark, nevermind the low water mark.).

> Actually, there is a bug. Sometimes when requesting work from a project that
> is disabled, BOINC then immediately requests 1 second of work from another
> project. Even if the queue is overfilled (above the highwater mark, nevermind
> the low water mark.).

If it's doing that it has a bigger problem than just the 1 second deal.

If the cache is overfilled (above the highwater mark) why would it have asked ANY PROJECT (even the disabled one) for work in the first place? It would seem the fact that it asked for work at all when above the highwater mark to be a bigger bug than the fact that it defaulted back to secondary project and asked for only 1 second of work.


<a> [/url]

> > Actually, there is a bug. Sometimes when requesting work from a project
> that
> > is disabled, BOINC then immediately requests 1 second of work from
> another
> > project. Even if the queue is overfilled (above the highwater mark,
> nevermind
> > the low water mark.).
>
> If it's doing that it has a bigger problem than just the 1 second deal.
>
> If the cache is overfilled (above the highwater mark) why would it have asked
> ANY PROJECT (even the disabled one) for work in the first place? It would
> seem the fact that it asked for work at all when above the highwater mark to
> be a bigger bug than the fact that it defaulted back to secondary project and
> asked for only 1 second of work.
>
The reason that it is requesting the one second has to do with multiple processor machines. I have a couple of dual 200MHz machines, these machines take two days to finish a WU. I have my cache set to 0.1 to 0.2 days. The typical way that these machines get work is to have a CPU run out of work. Note that at this time, the work remaining to do is usually greater than the high water mark, but there are not enough WUs to spread around to the CPUs available. Originally, that CPU was starved until the other WU was completed to below the low water mark. At this time BOINC requests 1 second of work from some project. The code does this by incrementing a variable someplace. The bug is that this variable is also incremented at times that it should not be incremented. This is clearly a case of a fix that was implemented too quickly causing other bugs.

You could always set the particular project to zero CPU time. =)

> You could always set the particular project to zero CPU time. =)
>
>

Yes, but if a project goes down suddenly (for instance a hardware failure) and won't be back up for, say, a week... How can you update your preferences to reflect the '0'?

A 'suspend' feature for individual projects would be VERY useful for situtations like this, so you wouldn't have to detatch the project, yet your other projects could keep on working. You could avoid hammering the down project until it came back up. (side effect of holding down all the 'red icon' producing error messages in the log too).
------------
KWSN-MajorKong
KWSN Forum Admin (retired)
http://www.kwsnforum.com

BOINC Beta tester

> Yes, but if a project goes down suddenly (for instance a hardware failure) and
> won't be back up for, say, a week... How can you update your preferences to
> reflect the '0'?

Not to mention that when a project is down (and you get the zeros for credit in the GUI) the resource share for that project is bumped back to the default 100.

This makes things a bit difficult.

<a> [/url]

> > Actually, there is a bug. Sometimes when requesting work from a project
> that
> > is disabled, BOINC then immediately requests 1 second of work from
> another
> > project. Even if the queue is overfilled (above the highwater mark,
> nevermind
> > the low water mark.).
>
> If it's doing that it has a bigger problem than just the 1 second deal.
>
> If the cache is overfilled (above the highwater mark) why would it have asked
> ANY PROJECT (even the disabled one) for work in the first place? It would
> seem the fact that it asked for work at all when above the highwater mark to
> be a bigger bug than the fact that it defaulted back to secondary project and
> asked for only 1 second of work.
>
>
It is not actually trying to get work. Once BOINC fails to connect to a project appoximately ten times it attempts redownload information from the master URL. The master fetch occurs independantly of the queue and does not automatically get work. The bug is that when the master fetch fails the client then requests 1 second of work from another project.

John Keck
BOINCing since 2002/12/08