BT 0.78

[jjwhalen]

Interesting is that this slowed state persists just as long as the BTs are connected to the particular client.

Maybe I've got it? I've just seen that the History tab's status bar constantly displays  "Updating..." without the usual 30 second countdown. I've seen something like this in the past when observing a task approaching its finish. A look into the Tasks tab: One FreeHal@home pseudo-nCi task was claiming being progressed to some couple of hundreds or thousands %...
I've aborted the task, updated project - as the task disappeared, BT was suddenly working as usual.. After downloading a new FreeHal@home task and starting it, in a while (around 141.756% seems to be enough) BT was sluggish again
Suspending the task seems to be enough to heal the sluggishness.

So, the idea was that maybe the high progress percentage triggers BT to go into "superfast history fetching" mode for a task approaching its end? The task contains sometimes no ETA, or "Estimated time remaining      -", sometimes just a few seconds...

BTW, what about displaying the weird progress percentage? BOINC Manager displays no (0.000%) Fraction done for the task, the same is true for BT's task properties window. But the Tasks tab often displays a strange value. Is it somehow computed? I suspect that the strange application generates no fraction_done, among other things it also generates no boinc_task_state.xml file in its slot...

I can confirm Pepo's observations (good catch, Peter!).  Newly downloaded FreeHAL WUs in (let's call it startup mode for want of a better term) appear to be gumming up the works.  The impossibly high Progress % seems to persist until the task reaches its first checkpoint, then changes to a credible 0.nnn% and increments at a believable rate.  Suspending project or disconnecting the affected computer appear to restore BT's normal update frequency immediately. 

I recall some discussion in this forum a few months ago about BT's algorithm for Progress %, but don't remember the details.  It looks like FreeHAL's "unique" way of doing business isn't appreciated by that process.  I also noticed that after aborting a FreeHAL task the displayed State changes to Aborted, but Progress % continues to increment for some time, possibly the rest of the current checkpoint interval, after which it changes to 100.000%.  Whether the task is actually still running after an abort command I don't know, but I wouldn't be surprised.

[fred]

A task that reports a time left close to 0 will result in a pretty fast re fetch.
But this shouldn't be a problem in 0.79 as it doesn't read all the data.
But anyway I will change the minimum time to 2 seconds.

[Beyond]

Interesting is that this slowed state persists just as long as the BTs are connected to the particular client.

Maybe I've got it? I've just seen that the History tab's status bar constantly displays  "Updating..." without the usual 30 second countdown. I've seen something like this in the past when observing a task approaching its finish. A look into the Tasks tab: One FreeHal@home pseudo-nCi task was claiming being progressed to some couple of hundreds or thousands %...
I've aborted the task, updated project - as the task disappeared, BT was suddenly working as usual.. After downloading a new FreeHal@home task and starting it, in a while (around 141.756% seems to be enough) BT was sluggish again
Suspending the task seems to be enough to heal the sluggishness.

So, the idea was that maybe the high progress percentage triggers BT to go into "superfast history fetching" mode for a task approaching its end? The task contains sometimes no ETA, or "Estimated time remaining      -", sometimes just a few seconds...

BTW, what about displaying the weird progress percentage? BOINC Manager displays no (0.000%) Fraction done for the task, the same is true for BT's task properties window. But the Tasks tab often displays a strange value. Is it somehow computed? I suspect that the strange application generates no fraction_done, among other things it also generates no boinc_task_state.xml file in its slot...

I'm also running FreeHAL on all 9 clients and can confirm when one or more nears completion BT starts locking up far more often.  The History tab behavior you describe also happens here.  When no FreeHAL WU is within 20% of completion the BT Tasks window updates much more reliably.

[fred]

I'm also running FreeHAL on all 9 clients and can confirm when one or more nears completion BT starts locking up far more often.  The History tab behavior you describe also happens here.  When no FreeHAL WU is within 20% of completion the BT Tasks window updates much more reliably.

V 0.80 will limit the time to 2 seconds, that should prevent any lockup. And with the Smart mode enabled the load should be far lower.
I will see if I can exclude FreeHAL.

[jjwhalen]

FreeHAL's new nci release 1.83 apparently has a fix for the weird progress indications at the start of a task, as described earlier in this thread by Pepo.  There's been some talk about it over on their user forum.  They were down most of yesterday, obviously implementing the changes.

At least the first 1.83 tasks I received this morning (California time) all showed a "normal" progress progression from startup.  Cautious optimism.

[Pepo]

FreeHAL's new nci release 1.83 apparently has a fix for the weird progress indications at the start of a task, as described earlier in this thread by Pepo.  There's been some talk about it over on their user forum.  They were down most of yesterday, obviously implementing the changes.

At least the first 1.83 tasks I received this morning (California time) all showed a "normal" progress progression from startup.  Cautious optimism.

I've observed both 1.83 and 1.84 tasks appearing to be "progressing as normal", but just now saw one 1.84 at 210% - the project team is either not able or has no intention to fix the weird progress behavior of their application

Luckily such weird progress is rare nowadays, much more applications behaved such weirdly in the past. A few years ago I'd be wishing to have an optional Progress% cell highlighting for debugging purposes for a task, which % is ± out of interval <0.000,99.999) or exactly at 100.000 but with non-zero remaining time. But FreeHal alone is not worth the additional work.