I noticed that my credit did not change for days. When I entered my account on climateprediction.net I saw that the last trickle happend at timestep 54010 in phase 2. Actually there is timestep 79500 and there are 2 trickles missing at 64812 and 75614. Is it possible to send the trickles manually? What can I do to avoid dataloss? (Boinc 4.19, hadsm3.4.04, Windows ME 4.90.3000 Build 3000)

---

What happens when you go to the Projects tab, right click on climateprediction.net and choose Update? Are there any error messages in the Messages tab at that point? Is this PC the one you are using to post this question?

Yes, this is the PC I used to post my question. When I click right on climateprediction.net in the projects tab there is no error message but:

climateprediction.net - 2005-04-16 19:43:29 - Sending request to scheduler: http://climateapps2.oucs.ox.ac.uk/cpdnboinc_cgi/cgi
climateprediction.net - 2005-04-16 19:43:33 - Scheduler RPC to http://climateapps2.oucs.ox.ac.uk/cpdnboinc_cgi/cgi succeeded

Obviously everything seems to be okay... Any idea?

---

> ...
> Obviously everything seems to be okay... Any idea?


You're right, everything seems okay. When that connection is successfully made, any unsent Trickles are sent up. (There is an issue if there are more than 40 Trickles in the queue, but that isn't your case.)

If you check Your Account, are the Trickles logged now?

Edit: By the way, there is no data loss if the Trickles go astray -- except for the End-of-Phase Trickles, they don't contain data, merely a way to inform the server that the run is still alive. No credit is lost; it will merely be delayed until the end of the run when full credit is awarded.

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

Thank you for your help. The latest trickle appeared in the list and the credit was immediately corrected.

---

Currently there seems to be a problem with the trickles though.

My trickles arrive at the scheduler, I can see that in the sched_request/reply (and no errors in stderr.txt) but they don't appear on the web page. No trickle files waiting to be sent in the CPDN project folder either.

The XML files say PH=3 and TS=108020 with fraction_done=0.808240 (which would match Trickle 58) but the web page says PH=3 TS=97218 and Trickle 57

So either just the web pages don't get updated (which wouldn't bother me too much) or the trickles disappeared in nirvana.

Afaik. if a trickle is missing, the finished model will not be accepted by the server and that would really bug me :-/

Result in question is 687541, the trickle has been sent about 2 hours ago, about midnight here which equals 22:00 UTC - usually trickles are not more than one hour delayed.

> Currently there seems to be a problem with the trickles though.
>
> My trickles arrive at the scheduler, I can see that in the sched_request/reply
> (and no errors in stderr.txt) but they don't appear on the web page. No
> trickle files waiting to be sent in the CPDN project folder either.
>
> The XML files say PH=3 and TS=108020 with fraction_done=0.808240 (which would
> match Trickle 58) but the web page says PH=3 TS=97218 and Trickle 57
>
> So either just the web pages don't get updated (which wouldn't bother me too
> much) or the trickles disappeared in nirvana.
>
> Afaik. if a trickle is missing, the finished model will not be accepted by the
> server and that would really bug me :-/
>
> Result in question is 687541, the trickle has been sent about 2 hours ago,
> about midnight here which equals 22:00 UTC - usually trickles are not more
> than one hour delayed.

Missing Trickles have nothing to do with the upload server accepting the five end-of-run .zip upload files.

The three end-of-Phase Trickles carry information used to display the graphs. The other 69 Trickles simply tell the server that the model is alive and to award a bit of credit. Missing Trickles would delay posting credit, nothing more. (Even if all Trickles were lost, full credit for the Run would be awarded when it finishes.)

As to not being up to date, it's no big deal (for me at least) -- I have five P4's running ten Models. They normally upload Trickles for each Model from about 7.5 to 9.3 hours. So, in less than ten hours, in normal conditions, all Models are heard from. Not so now. Last Trickle was posted @ 14:39:26, about ten hours ago.

Probably some glitch in the database or communications. It'll be sorted out sooner or later.

Hang in there, Ananas. When Tolu gets help, supposedly next month, things should run smoother. ... I hope.

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

I knew that the trickles themselves do not carry any project information but found a thread that says the server needs them to assure data consistency :

http://www.climateprediction.net/board/viewtopic.php?t=2811


So I hope you're right and the scheduler did receive the trickles somehow - I was already thinking about a good way to switch web access off for the PCs running CPDN without a monitor attached :-/


I'm not so worried about the credits but I had quite some problems with my first models and I planned to have listed more full runs than crashes soon.

Thanks for your answer and have a nice evening :-)

Good night *yawn*
Volker

edit : If some trickle got lost, I can still make a few with a text editor ;-)

---

I'm getting the same problem, and so are <a href="http://www.climateprediction.net/board/viewtopic.php?p=26183">others</a>. Another server problem.

---

&gt; ...
&gt; edit : If some trickle got lost, I can still make a few with a text editor ;-)


Just had to do that, trickle 50 and 51 have been missing in a model that just arrived at trickle 71. It accepted those trickles although they have been quite late.

So it seems, if a full run doesn't change to state "Over / Success / Done", one can add missing trickles :-)

---

I hope it wasn't wrong to manually create the trickles that have been lost when the server was acting up.


It has 72 trickles now, BOINC reported it but it doesn't go to the final state.

Result in question is :

<a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=631579">result.php?resultid=631579</a>


Any ideas? I don't like the idea of having trashed one more result :´(

---

&gt; Any ideas? I don't like the idea of having trashed one more result :´(

The result state for completed jobs doesn't change until your first scheduler request after the result files have been uploaded. That host hasn't made any trickles since the last one for your completed model, but you could force the state to change before your next trickle by doing a manual update.

---

"The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer

I made contact with manual "update" short after the final trickle.

The next model (2whu_200156792_0) did trickle at 9am (CET+DST) which is about 9 hours ago but the trickle didn't show up yet. The reply says "message_ack/" so the trickle must have arrived.

Maybe there is a server problem again.


p.s.: 9am here would be 7am on the CPDN server


<b>The server IS loosing data :-(</b>

My result 767767 trickled #5 now but #4 still didn't appear. Result 631579 is still Unknown/New although it is over and BOINC did report it to the server.

I checked for message_ack/ and it had been there so it was not a communication problem.

---

I seem to have lost another trickle, on the just started model.
Doesn't seem much point in 'talking' to the server daily, so I'll save then until there's a post by admin that the problem's been fixed.

Les

The weird thing is that the SQL server says there isn't a problem.

- For the trickles I got an acknowledge

- Same for the reported finished model

- When I tried to change the preferences it said "has been changed"

In all 3 cases the database update seems to have failed but there seems not to be an error return code or status code that would make the programs repeat the update or insert attempt.

I don't think it is a BOINC or CPDN software problem, sounds more like a MySQL or PHP bug or faulty database contents.

---

To illustrate the missing trickles problem, my results: <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/show_user.php?userid=3230"> here</a>.
1 CPU machine has done all right, sending trickles one at a time.
HT machine has got missing trickles.
I've got an ADSL connection I turn off when I'm not around.
As a result it may send a few trickles simultaneously.
Could that be the problem?

---

Forum search Site search

Usually not a problem to send multiple trickles, as long as it's not really many of them. If you hit the limit, it will not work at all anymore, so if some got sent it isn't this limit.

---

Probably related to the loss of data, I currently get this error when I try to access my show_user.php :

can't read ../cache/43/show_user.php_userid%3D27928; lastmod 1115560201

Several more trickles missing now too :´(


edit : I think the scheduler should be shut down until the problem has been fixed.

---