The figure 8 way is to first suspend the network, but that suspending tasks/projects leads to immediate fetches of other *active* projects is a feature that got killed about version 5.6, maybe earlier. When your client buffer is full, including the still counted suspended work, it's full.

--//--

...but that suspending tasks/projects leads to immediate fetches of other *active* projects is a feature that got killed about version 5.6, maybe earlier.

I'd say later, because 5.10.45 still does it ;-)

Gruß,
Gundolf

Hard to believe, but maybe we're talking different things. There was a time that suspending tasks were not counted for the cache and that's been gone a very long time. If suspending tasks, when technically the cache is full, then fetching work from other projects to exceed the total value of cache, I'd be surprised. I've had a task for longer in suspend [the last task], not the project, and BT tells me the sum of all work is still what I've got set, and never requesting more unless that projected total [of all projects] drops below the set value. Same observation for 6.10 and 6.12.

--//--

Hard to believe, but maybe we're talking different things.

I'm talking of the BOINC version (but BoincTasks is in the 1.2x range).

...I've had a task for longer in suspend [the last task], not the project...

There's my mistake: I had suspended a project, not its tasks.

But you've also been only 'half' right:

that suspending tasks/projects leads to immediate fetches of other *active* projects is a feature that got killed about version 5.6

(emphasis by me) since suspending a project definitely causes overfetch with 5.10.45 ;-)

Gruß,
Gundolf

Hard to believe, but maybe we're talking different things. There was a time that suspending tasks were not counted for the cache and that's been gone a very long time. If suspending tasks, when technically the cache is full, then fetching work from other projects to exceed the total value of cache, I'd be surprised. I've had a task for longer in suspend [the last task], not the project, and BT tells me the sum of all work is still what I've got set, and never requesting more unless that projected total [of all projects] drops below the set value. Same observation for 6.10 and 6.12.

--//--

I'm sure that suspended projects/tasks aren't counted towards the cache (in Boinc 6.10.58 and 6.12.41), i tend to get a small stash of CPDN tasks for a rainy day,
then suspend them all but one, and fill up with tasks from Seti and Seti Beta, i've been caught out a few times when i've not had NNT set at a project, and i've suspended another project,
and suddenly had extra work that i hadn't intended to get.

Claggy

Hard to believe, but maybe we're talking different things. There was a time that suspending tasks were not counted for the cache and that's been gone a very long time. If suspending tasks, when technically the cache is full, then fetching work from other projects to exceed the total value of cache, I'd be surprised. I've had a task for longer in suspend [the last task], not the project, and BT tells me the sum of all work is still what I've got set, and never requesting more unless that projected total [of all projects] drops below the set value. Same observation for 6.10 and 6.12.

--//--

I'm sure that suspended projects/tasks aren't counted towards the cache (in Boinc 6.10.58 and 6.12.41), i tend to get a small stash of CPDN tasks for a rainy day,
then suspend them all but one, and fill up with tasks from Seti and Seti Beta, i've been caught out a few times when i've not had NNT set at a project, and i've suspended another project,
and suddenly had extra work that i hadn't intended to get.

Claggy

And I got caught the same way when suspending tasks in BOINC v6.13.10

And I got caught the same way when suspending tasks in BOINC v6.13.10

Ah, you as well? Got that as well, a whole cache full of Enigma's and Einstein's when I temporarily suspended the tasks not Albert, when trying to 'force' the OpenCL app to start up.

---

And I got caught the same way when suspending tasks in BOINC v6.13.10

Ah, you as well? Got that as well, a whole cache full of Enigma's and Einstein's when I temporarily suspended the tasks not Albert, when trying to 'force' the OpenCL app to start up.

In my case, I got a dollop of NumberFields, which - to add to the stress - the server had allocated without reference to the client's current DCF of something over 7.5: two and three quarter million unexpected seconds of work....

See BOINC Dev mailing list.