I finished and uploaded model 346646. I did not notice any errors from BOINC. But CPDN says it's still "in progress" and the P3 graph is unavailable. I have saved the results. Did CPDN get everything to make this a scientificly successfull model?

Regarding the graph, the page may not have been updated yet on the server (it can take a few hours).
When I finish a model it stays in 'work' in the BOINC GUI till the next time it connects (at the first trickle from the next model). That's when the result changes to completed.
What do things look like for you in the GUI? Anything left in 'transfer'?

---

Forum search Site search

This model uploaded on Jan 31
The "in process" I'm referring to is here:
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/results.php?hostid=71868

The upload occurred with BOINC 4.13

---

> This model uploaded on Jan 31
> The "in process" I'm referring to is here:
> http://climateapps2.oucs.ox.ac.uk/cpdnboinc/results.php?hostid=71868
>
> The upload occurred with BOINC 4.13
>
It looks like the last trickle of phase 3 was not recorded as uploaded. When that happens, the graphs aren't generated.

Has the PC communicated with the server since Jan 31st? It looks like no trickles from that host since the last 71st of your model. What happens if you restart BOINC on that PC?

I'm a dial-up user. Running Linux (CLI interface). Often, there is no connection when the client wants to communicate with the server, so the client *schedules* another communication attempt at a future time.

Something (I don't know what) went wrong, such that client uploads of trickles were no longer successful. (Instead, a message was typed by the client: "No server responding".) But the client dutifully kept __scheduling__ "contact" times, the last of which happened to be a week in the future.

I let the workunit complete phase 3. When my computer went idle (it couldn't successfully download new work, either) I manually restarted with '-update prefs' -- that resulted in the five results files being uploaded.

When I couldn't "brute force" the client to upload its stacked-up trickles (it still gave me "No server responding"), I detached from the CPDN project, then re-attached. That threw away all previous trickles, etc., but re-downloaded the hadsm3 stuff. The client is now running (without any problems, as far as I can tell) on a new workunit.

--------

My reason for posting is: The present BOINC design places undue importance on the transmission of trickles. Even if I did not have the upload problem, I would have had to wait a WEEK for the servers to finally mark the previous unit as 'complete'. PLEASE, PLEASE, PLEASE, when the final results files have been uploaded, do the processing for 'complete' then and there. As my case showed, there may __never__ occur that "follow-on" trickle on which the present BOINC design depends.
.

If your connection isn't always on. In the BOINC GUI file menu:
Check: disable network access
(and uncheck it when you are connected and want to send trickles)
That way, you won't get any communications error messages.

If you have messages like deferring communications for...
In the projects tab:
Select the project you want
Right-click
Select update

---

Forum search Site search

> It looks like the last trickle of phase 3 was not recorded as uploaded. When
> that happens, the graphs aren't generated.
>
> Has the PC communicated with the server since Jan 31st? It looks like no
> trickles from that host since the last 71st of your model. What happens if
> you restart BOINC on that PC?
>

This is my slowest PC at ~6.6 sec/T. I decided it should switch to smaller work units. I set the "leave at least this much free space" setting to 99GB to prevent another model download. After BOINC uploaded all the files it wanted, I upgraded to 4.19 and attached to Predictor and LHC. It is happily running Predictor and will run LHC when it is back.

Could the leave free space option have prevented a critial file from uploading?

mickus says something about a "follow up trickle." What is this?

---

Loch Dhu asks: Mikus says something about a "follow on trickle". What is this?

I was referring to the post on 2 Feb by 'mavau', who says:
> When I finish a model it stays in 'work' in the BOINC GUI till the next time
> it connects (at the first trickle from the next model). That's when the result
> changes to completed.

[By the way, in my case because the '-attach_project' wiped out the client's knowledge of the previous workunit, that model's status at the server NEVER changed to completed, although I've now uploaded the first trickles from the next model.]


Comment to the post on 3 Feb by 'mavau': I'm running Linux (CLI interface). I have __no__ GUI file menu in which to check/uncheck settings. And on Linux, the ONLY (as far as I know) update option I have is '-update_prefs' -- the trickle update schedule seems to be unaffected by that option.
.