Column for "Uncheckpointed CPU time"

[Pepo]

If ever such column will be added in the Tasks tab, the displayed value could be the difference ("CPU time at last checkpoint" - "CPU time"). For strategic games like "do I shut down the client now, or should I wait a bit more..."
(BOINC client devs are not willing to add something like "Pause each running app after reaching a checkpoint".)

[jjwhalen]

If ever such column will be added in the Tasks tab, the displayed value could be the difference ("CPU time at last checkpoint" - "CPU time"). For strategic games like "do I shut down the client now, or should I wait a bit more..."
(BOINC client devs are not willing to add something like "Pause each running app after reaching a checkpoint".)

This seems do-able, since the information is already available in task properties.  I know which of my subprojects have long checkpoint intervals (or no checkpointing at all) and can judge when to shut down a client or reboot.  But for someone with many hosts...very helpful.  I could also envision warning colors for things like >1hr since last checkpoint.

[fred]

Added to the list.

[jjwhalen]

Added to the list.

Good call Pepo

[Pepo]

Added to the list.

Good call Pepo

I'm glad to be able to propose anything useful.

[fred]

What I made is a column that shows the time period, past the last checkpoint.

00:02:00 the last checkpoint was 2 minutes ago.

Is the (cputime - checkpoint time) / (CPU% / 100)

[Pepo]

00:02:00 the last checkpoint was 2 minutes ago.
Is the (cputime - checkpoint time) / (CPU% / 100)

That is "sort of" "Elapsed time since the last checkpoint". But which CPU% average?? Long-term? May have no meaning for just the period since the last checkpoint. Overall CPU load and throttling may vary with time. Short-term? May have no meaning for the whole period since the last checkpoint. For the same reasons.
I believe pure CPU time difference would tell more... or less... Actually exactly the questionable value.

[John McLeod VII]

I have a couple of minor suggestions for the time since last checkpoint column.

1)  Could the color in the column be green even if the color for the line is red or yellow for some reason (assuming that the checkpoint is recent).

2)  Could we add yellow to mean that the checkpoint has not occurred within the last 2* minimum checkpoint interval?  (same color request as 1 - yellow even if line is red if column is in this state).

[fred]

I have a couple of minor suggestions for the time since last checkpoint column.

1)  Could the color in the column be green even if the color for the line is red or yellow for some reason (assuming that the checkpoint is recent).

2)  Could we add yellow to mean that the checkpoint has not occurred within the last 2* minimum checkpoint interval?  (same color request as 1 - yellow even if line is red if column is in this state).

1) I made the checkpoint item a different color so that would be possible.
2) Could do that, involves reading the interval but that's not too difficult.

I put it on the list.

[jjwhalen]

I have a couple of minor suggestions for the time since last checkpoint column.

1)  Could the color in the column be green even if the color for the line is red or yellow for some reason (assuming that the checkpoint is recent).

2)  Could we add yellow to mean that the checkpoint has not occurred within the last 2* minimum checkpoint interval?  (same color request as 1 - yellow even if line is red if column is in this state).

1) I made the checkpoint item a different color so that would be possible.
2) Could do that, involves reading the interval but that's not too difficult.

I put it on the list.

I appreciate the quick implementation of Pepo's Checkpoint column & the warning.

1) I'd request that any checkpoint warning be made checkbox-optional in Settings.  (I should've mentioned that when I suggested it originally )

2) Separate warning colors for CPU & GPU might also be handy.  (Another "shoulda' thought of that before" )

3) Would it be possible to generate an exclusion list of "projects or applications I don't want to get checkpoint warnings about"   Which I suppose raises the question of exclusion lists for other things in BT, including the GUI to enter exclusion items as well as to go back later and review/edit/delete...hmm   Maybe that's one of those 'bottom of the wishlist' things

[fred]

I appreciate the quick implementation of Pepo's Checkpoint column & the warning.

1) I'd request that any checkpoint warning be made checkbox-optional in Settings.  (I should've mentioned that when I suggested it originally )

2) Separate warning colors for CPU & GPU might also be handy.  (Another "shoulda' thought of that before" )

3) Would it be possible to generate an exclusion list of "projects or applications I don't want to get checkpoint warnings about"   Which I suppose raises the question of exclusion lists for other things in BT, including the GUI to enter exclusion items as well as to go back later and review/edit/delete...hmm   Maybe that's one of those 'bottom of the wishlist' things

You want everything.
Lets see.

[fred]

2)  Could we add yellow to mean that the checkpoint has not occurred within the last 2* minimum checkpoint interval?  (same color request as 1 - yellow even if line is red if column is in this state).

A quick check: What do mean by minimum checkpoint interval.
Extra->BOINC preference -> Disk -> Write to disk at most every.

[Pepo]

2)  Could we add yellow to mean that the checkpoint has not occurred within the last 2* minimum checkpoint interval?  (same color request as 1 - yellow even if line is red if column is in this state).

A quick check: What do mean by minimum checkpoint interval.
Extra->BOINC preference -> Disk -> Write to disk at most every.

I guess exactly this is it.

[jjwhalen]

2)  Could we add yellow to mean that the checkpoint has not occurred within the last 2* minimum checkpoint interval?  (same color request as 1 - yellow even if line is red if column is in this state).

A quick check: What do mean by minimum checkpoint interval.
Extra->BOINC preference -> Disk -> Write to disk at most every.

That's what I took John's suggestion to mean, and it sounds reasonable.  FYI earlier in this thread when I said

I could also envision warning colors for things like >1hr since last checkpoint

I was thinking mostly of the wrapper-apps, which can have some very strange checkpointing.  For example Yoyo's Cruncher OGR often goes 2~3 hours between checkpoints, and his EVO runs for hours and doesn't checkpoint at all.

[John McLeod VII]

I was after a two step warning.

Yellow meaning > 2 * "write to disk at most"

and

Red meaning > 1 hour (or possibly user specified).