i\'m afraid that WU with ID 109 DIED after 42 trickles over weekend (when i wasn\'t present) and client started processing another WU [very sad]
It says \'exit code -5\' [why negative numbers?]
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=109

BOINC forum is not very readable and search engine doesn\'t seem to work so i\'m unable to tell if there has been such problem before...anyone already seen \'exit code -5\'?

Is there a list of error/exit codes?

I assume it uploaded the results & your yabsd.out, I'll go lookup on the servers tomorrow and see what happened.

Hi, Honza,

This was saved from days gone by. Cryptic but useful.

Jim


Posted: 24 Jun 2004 22:27:09 UTC
 by Knightrider

Full file and ,maybe, updates can be found here:
 
 http://boinc.berkeley.edu/error_numbers.h
 
 #define BOINC_SUCCESS 0
 #define ERR_SELECT -100
 #define ERR_MALLOC -101
 #define ERR_READ -102
 #define ERR_WRITE -103
 #define ERR_FREAD -104
 #define ERR_FWRITE -105
 #define ERR_IO -106
 #define ERR_CONNECT -107
 #define ERR_FOPEN -108
 #define ERR_RENAME -109
 #define ERR_UNLINK -110
 #define ERR_OPENDIR -111
 #define ERR_XML_PARSE -112
 // Unexpected XML tag or XML format
 #define ERR_GETHOSTBYNAME -113
 // Couldn't resolve hostname
 #define ERR_GIVEUP_DOWNLOAD -114
 // too much time has elapsed without progress on file xfer,
 // or we got a 404 (not found) status code from server
 #define ERR_GIVEUP_UPLOAD -115
 #define ERR_NULL -116
 // unexpected NULL pointer
 #define ERR_NEG -117
 // unexpected negative value
 #define ERR_BUFFER_OVERFLOW -118
 // caught buffer overflow
 #define ERR_MD5_FAILED -119
 // MD5 checksum failed for a file
 #define ERR_RSA_FAILED -120
 // RSA key check failed for a file
 #define ERR_OPEN -121
 #define ERR_DUP2 -122
 #define ERR_NO_SIGNATURE -123
 #define ERR_THREAD -124
 // Error creating a thread
 #define ERR_SIGNAL_CATCH -125
 #define ERR_QUIT_REQUEST -126
 // The app exited due to user request and should be restarted later
 #define ERR_UPLOAD_TRANSIENT -127
 #define ERR_UPLOAD_PERMANENT -128
 #define ERR_IDLE_PERIOD -129
 // can't start work because of user prefs
 #define ERR_ALREADY_ATTACHED -130
 #define ERR_FILE_TOO_BIG -131
 // an output file was bigger than max_nbytes
 #define ERR_GETRUSAGE -132
 // getrusage failed
 #define ERR_BENCHMARK_FAILED -133
 #define ERR_BAD_HEX_FORMAT -134
 // hex-format key data is bad
 #define ERR_USER_REJECTED -135
 // user rejected executable file
 #define ERR_DB_NOT_FOUND -136
 // no rows found in lookup()
 #define ERR_DB_NOT_UNIQUE -137
 // not unique in lookup()
 #define ERR_DB_CANT_CONNECT -138
 #define ERR_GETS -139
 // gets() or fgets()
 #define ERR_SCANF -140
 // scanf() or fscanf()
 #define ERR_STRCHR -141
 #define ERR_STRSTR -142
 #define ERR_READDIR -143
 #define ERR_SHMGET -144
 #define ERR_SHMCTL -145
 #define ERR_SHMAT -146
 #define ERR_FORK -147
 #define ERR_EXEC -148
 #define ERR_NOT_EXITED -149
 // a process didn't exit that was supposed to
 #define ERR_NOT_IMPLEMENTED -150
 // a system call not implemented on this platform
 #define ERR_GETHOSTNAME -151
 #define ERR_NETOPEN -152
 #define ERR_SOCKET -153
 #define ERR_FCNTL -154
 #define ERR_AUTHENTICATOR -155
 // scheduler request host ID doesn't match authenticator
 #define ERR_SCHED_SHMEM -156
 // sched shmem has bad contents
 #define ERR_ASYNCSELECT -157
 #define ERR_BAD_RESULT_STATE -158
 #define ERR_DB_CANT_INIT -159
 #define ERR_NOT_UNIQUE -160
 // state files had redundant entries
 #define ERR_NOT_FOUND -161
 // inconsistent client state
 #define ERR_NO_EXIT_STATUS -162
 // exit_status not found in scheduler request
 #define ERR_FILE_MISSING -163
 #define ERR_NESTED_UNHANDLED_EXCEPTION_DETECTED -164
 #define ERR_SEMGET -165
 #define ERR_SEMCTL -166
 #define ERR_SEMOP -167
 #define ERR_FTOK -168
 #define ERR_SOCKS_UNKNOWN_FAILURE -169
 #define ERR_SOCKS_REQUEST_FAILED -170
 #define ERR_SOCKS_BAD_USER_PASS -171
 #define ERR_SOCKS_UNKNOWN_SERVER_VERSION -172
 #define ERR_SOCKS_UNSUPPORTED -173
 #define ERR_SOCKS_CANT_REACH_HOST -174
 #define ERR_SOCKS_CONN_REFUSED -175
 #define ERR_TIMER_INIT -176
 #define ERR_RSC_LIMIT_EXCEEDED -177
 #define ERR_INVALID_PARAM -178
 #define ERR_SIGNAL_OP -179
 #define ERR_BIND -180
 #define ERR_LISTEN -181
 #define ERR_TIMEOUT -182
 #define ERR_PROJECT_DOWN -183
 
________________________________________________
It is impossible to enjoy idling thoroughly unless one has plenty of work to do.
-- Jerome K. Jerome (1859, 1927)

well -5 is my catch all "model crashed" error, for Honza's run I found this from his yabsd.out on the upload server, so it was definitely from a calculation gone awry. I may put in more "forceful" checking in the Fortran compiler, which would slow things 5-10% but would probably be better to avoid these kinds of errors.

--------------------------------------------------------------------

DAS- LSP_FORM- QCL and DELTA not updated: NaN 1800.000
NaN
DAS- LSP_FORM- QCL and DELTA not updated: NaN 1800.000
NaN
QT_POS : Mass weighted QT summed over level 4
was negative. WARNING: QT not conserved
1 points were -ve and the scaling factor has been reset to 1
QT_POS : Mass weighted QT summed over level 5
was negative. WARNING: QT not conserved
1 points were -ve and the scaling factor has been reset to 1
QT_POS : Mass weighted QT summed over level 6
was negative. WARNING: QT not conserved
10 points were -ve and the scaling factor has been reset to 1
NEGATIVE THETA AT POINT 5638 LEVEL 1
NEGATIVE THETA AT POINT 5638 LEVEL 2
NEGATIVE THETA AT POINT 5638 LEVEL 3
NEGATIVE THETA AT POINT 5638 LEVEL 4
NEGATIVE THETA AT POINT 5639 LEVEL 5
NEGATIVE THETA AT POINT 5831 LEVEL 6
NEGATIVE THETA AT POINT 6023 LEVEL 7
*********************************************************************************
Model aborted with error code - 1 Routine and message:-
ATM_DYN : NEGATIVE THETA DETECTED.
*********************************************************************************
IN UMSHELL 12
1

Thank for looking into it, Carl.
As i understand it, it's kind of extreme climate but not in term on temperature.
btw, is there {uder BOINC) also kind of extreme climate -> 3x rewind to recalculate scenario?

This machine finished THC model so i decided to let Dekora machine continue on BOINC - pretty stable computer i guess.

> well -5 is my catch all "model crashed" error, for Honza's run I found this
> from his yabsd.out on the upload server, so it was definitely from a
> calculation gone awry. I may put in more "forceful" checking in the Fortran
> compiler, which would slow things 5-10% but would probably be better to avoid
> these kinds of errors.
>
> --------------------------------------------------------------------

Oh no, no again.
Have been running two WUs overnight and WU 11092 (after 1 trickle) and 11093 (before trickled) are dead; again exit code -5
WU 5985 still holding - now 7 trickles.

I realized that there are two potencial issues on this box:
1) slightly overcloked (3 -> 3.2), SuperPi on 4M just fine and never experienced unexpected restart not model crash under classic; might try to let it on standard 3.0
2) BOINC client, resp. whole directory was actually moved from another machine. I wonder is that makes any difference since also all is just fine (e.g. GUI, client-server communication etc.). I fact, i arise question of possible migrating form one machine to another (under same OS of cause).

I decided to restart, play nice to it (standard 3GHz) and let only one model running to see.
Any comments on 2)?

perhaps CPDN/BOINC is very sensitive to overclocking? I suppose I could compile with more rigorous floating point checks etc. Perhaps since we use Intel compilers now they are "less forgiving" of overclocking (and AMDs!)

> perhaps CPDN/BOINC is very sensitive to overclocking? I suppose I could
> compile with more rigorous floating point checks etc. Perhaps since we use
> Intel compilers now they are "less forgiving" of overclocking (and AMDs!)
>
Hmm, perhaps.

Still how about migration CPDN BOINC model from one computer to another?
Is there gonna be new build on 26th?

> Still how about migration CPDN BOINC model from one computer to another?
> Is there gonna be new build on 26th?

This isn't possible with WUs for other BOINC projects, so I suspect the same is true with CPDN.

<a><img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=18"></a>

&gt; well -5 is my catch all "model crashed" error, for Honza's run I found this
&gt; from his yabsd.out on the upload server, so it was definitely from a
&gt; calculation gone awry. I may put in more "forceful" checking in the Fortran
&gt; compiler, which would slow things 5-10% but would probably be better to avoid
&gt; these kinds of errors.
&gt;

Hi, Carl,

Do you have stats on % of OC'd machines vs. total now Trickling in "classic" CPDN?

Can you run a simulation of the consequences for the project of slowing everyone 5-10% to save an unknown % of failed Models on whatever % of OC'd machines?

Jim
________________________________________________
It is impossible to enjoy idling thoroughly unless one has plenty of work to do.
-- Jerome K. Jerome (1859, 1927)

&gt; &gt; Still how about migration CPDN BOINC model from one computer to another?
&gt;
&gt; This isn't possible with WUs for other BOINC projects, so I suspect the same
&gt; is true with CPDN.
&gt;
Well, if it isn't possible, there should be check before running WU that it is on the same computer (ID) and prevent it from running.

My migration (moving whole BOINC folder to another computer) worked with no glitch so far - no complaining from BOINC, just running benchmarks, Machine description updated. New model is doing good (16%); the old one went almost to Phase 3...
btw, i wan't had such problem with Alpha.

&gt; well -5 is my catch all "model crashed" error, for Honza's run I found this
&gt; from his yabsd.out on the upload server, so it was definitely from a
&gt; calculation gone awry. I may put in more "forceful" checking in the Fortran
&gt; compiler, which would slow things 5-10% but would probably be better to avoid
&gt; these kinds of errors.

Excuse me butting in but that seems a bit arse-about-face to me. You're considering reducing the processing speed by 5-10% for everybody because some people have made the concious decision to overlock their machine (an action that often causes problems) and are now seeing errors.

If I understand your comment you are just considering adding additional checks to the calculations so, unless I am missing the point, overclocked machines would still fail - you'd just have a better idea of when they failed?

It just seems a bit pointless to me. If someone overclocks then the onus should be on them to make their machine compatible with the project and not the other way round.

Ian

&gt; Excuse me butting in but that seems a bit arse-about-face to me. You're
&gt; considering reducing the processing speed by 5-10% for everybody because some
&gt; people have made the concious decision to overlock their machine (an action
&gt; that often causes problems) and are now seeing errors.
&gt;
&gt; If I understand your comment you are just considering adding additional checks
&gt; to the calculations so, unless I am missing the point, overclocked machines
&gt; would still fail - you'd just have a better idea of when they failed?
&gt;
&gt; It just seems a bit pointless to me. If someone overclocks then the onus
&gt; should be on them to make their machine compatible with the project and not
&gt; the other way round.
&gt;
&gt; Ian
&gt;

Hi Ian,

thanks for clear comments. I agree with you that client doesn't have to be adjusted to overclockers needs but to be stable as it is on 'factory settings' and not the other way round.

Let's calm down and have a look again:
note that this machine is VERY stable (regardless slight overclocking i havent found ANY problem so far; stability tests incluede) and completed some amount of standard and THC slowdown models.
My question (still not answered) was also about migration of on-going model - together with 'exit code -5'...
Also note that i'm not the only one with 'exit code -5' so connecting it apriory with decent overclocking is a not precedent.

Statisticaly: how do you explain model going toward Phase 3 and then crashed and some models crashes after couple of hours just in beginning?
Climate: some model crash due to paramteters
Beta: note that this is still beta - client and new BOINC 4.0x as well.

I'm not sure if you have been there during Alpha stage or had opportunity to talk to CPDN team in face... i can asure you that their are doing very good job. Also note that compiling 500k+ source code such as climate model is very sensitive (unlike SETI etc.) and that were many problems due to different compilers and switches - not only for Win.
Well, it also seems that AMD CPUs tends to loose on Intel P4 (unlike under classic CPDN as 1+year experience shows) due to compilers and it's switches. AMD with some switches tended to be faster and, yes - unstable...There must be decent balance of optimalization for performance and stability.

My point is that there are many factors that impact (in)stability...
What would you say if i report another crash under 'factory settings'?
What woud you say if there are more and more people reporting this problem?

This (and other 3 boxes i'm running at home) have showed a good proof of stability and compatability during more than a year running 24/7 classic CPDN since its early beta.

&gt; My question (still not answered) was also about migration of on-going model -
&gt; together with 'exit code -5'...

Well, I'll try again...

One of the new "features" BOINC has in place to stop credit cheating, (in relation to S@H, CPDN might not apply) is to only allow a WU to be reported by the machine that downloaded it.

Whether or not this caused your "exit code -5", I don't know. Were any trickles successfully reported after the migration?

<a><img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=18"></a>

Hi Honza,


&gt; Let's calm down and have a look again:

Whoops, did I come across too strong in my comments?. I certainly didn't intend to and apologise if it appeared that steam was coming out of my ears :-)


&gt; My point is that there are many factors that impact (in)stability...

OK, but Carl et al have written a program that, given a set of input parameters, will produce a consistent result - there are, or should not be, any compiler switches that will change that.

There are however external causes, for example overclocking, heat, processor differences &amp; cosmic rays randomly flipping bits, that _may_ cause the program to fail or, probably more seriously, produce incorrect results.

My point was just that it is difficult, if not impossible, to allow for that within the program without an excessive amount of error checking code [1]. The onus should be on the person running to program to make sure the platform is stable. Crashes on overclocked machines, if normally clocked machines do not show errors, would tend indicate to me that BOINC CPDN just stresses overclocked processors too much - especially if they are P4s with HT.

FWIW, I've been running the old CPDN client since the original beta test and BOINC for the last couple of weeks with no crashes at all.

Ian

[1] As a Smalltalk programmer who has never used Fortran I have no idea what compilation options are available these days.

&gt; One of the new "features" BOINC has in place to stop credit cheating, (in
&gt; relation to S@H, CPDN might not apply) is to only allow a WU to be reported by
&gt; the machine that downloaded it.
&gt;
&gt; Whether or not this caused your "exit code -5", I don't know. Were any
&gt; trickles successfully reported after the migration?
&gt;

Yes, WU ID 109 returned 42 trickles, most of them (including Post-phase 1 processing if i remember right) after migration.

Note that another 2 WUs 'exited' AFTER migration when 2 CPUs used (P4 HT).

It might have been with connection of prolonged heat (30+C over a week with no rain at all) and other consequences. I have checked CPU temp again and it's 63C; not so bad should be better i guess...others P4 are up to 60C.

@ Ian,

after rereading Andrew's updated post there (http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=206) i tend more to ascribe the 'exit code -5' problem to extreme heat in the room where affected conputer was. But unusual thing is that it was during dark night which confuses me. Will check CPU fan etc...
Thanks for mentioning other external causes - now i see that you are pretty familiar with issue.

I remember some note on Open Date about result of idenatical model on different machines (e.g. re-check) but can't remember exact match in % (was it 95% ?). Somebody will hopefully fill-in the gap or perhaps Open-Day archive will help there. I'm quite sure it was 100% match.

Hey,

now it shot me with exit code -5

First let me tell you, no, this computer is NOT overclocked !

This is the machine <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/show_host_detail.php?hostid=1134">http://climateapps2.oucs.ox.ac.uk/cpdnboinc/show_host_detail.php?hostid=1134</a>

It's a DELL Latitude C840 notebook, P4 M, 2GHz with 512 MB RAM, running WIN XP Prof SP1

The box is running since more than 1 year in BOINC-BETA, Seti/Boinc, Predictor and, since some days, CPDN. Nothing unusual happened in this time.

The only thing: In trickles, I saw, it must have crunched very fast for an 2 GHz P4, because it needed 3,17 to 3,29 s/TS. (Now I know, why I have paid so much ;-) )

Today, 2 hours before the exit code -5 happened, I have detached Seti on this box. I think, it should have nothing to do with it.

This is the result, that crashed now: <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=7721">http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=7721</a>

I can not recover, what happened in which timeline afterwards. When I recognized, what had happened, the firewall (on the notebook) had crashed, boinc-cc had crashed and the box had downloaded several WUs and crashed them.

Fortunately, the server stopped the box with the message "No work, Daily Quota exceeded".

I have made a copy of the whole BOINC_directory, so, if you need any information, any file sent, let me know


<a href="http://www.boinc.dk/index.php?page=user_statistics&amp;project=cpdn&amp;userid=34"><img src="http://34.cpdn.sig.boinc.dk?142"></a><img src="http://75.pah.sig.boinc.dk?142"><img src="http://4238.sah.sig.boinc.dk?142">

Hi Yeti,

i know how you feel now. My first BOINC model crashed after 42 trickles.
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/trickle.php?resultid=109

Since this is another model crashed reported on notebook - which tends to overheat - i assume i should have been due to overall overheat (in room e.g.) -&gt; system overheat. Well, mine crashed during night on a P4 desktop?!
Can you provide some estimates on room teperature and/or CPU temperature?

All models crashes with 'exit code -5' are P4 AFAIR.

The positive is that server has some Quota to send nn models a day and prevents "Short Run Maniac" phenomenon. Good idea.

&gt; Hey,
&gt;
&gt; now it shot me with exit code -5
&gt;
&gt; First let me tell you, no, this computer is NOT overclocked !
&gt;
&gt; This is the machine <a> href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/show_host_detail.php?hostid=1134"&gt;http://climateapps2.oucs.ox.ac.uk/cpdnboinc/show_host_detail.php?hostid=1134</a>
&gt;
&gt; It's a DELL Latitude C840 notebook, P4 M, 2GHz with 512 MB RAM, running WIN XP
&gt; Prof SP1
&gt;
&gt; The box is running since more than 1 year in BOINC-BETA, Seti/Boinc, Predictor
&gt; and, since some days, CPDN. Nothing unusual happened in this time.
&gt;
&gt; The only thing: In trickles, I saw, it must have crunched very fast for an 2
&gt; GHz P4, because it needed 3,17 to 3,29 s/TS. (Now I know, why I have paid so
&gt; much ;-) )
&gt;
&gt; Today, 2 hours before the exit code -5 happened, I have detached Seti on this
&gt; box. I think, it should have nothing to do with it.
&gt;
&gt; This is the result, that crashed now: <a> href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=7721"&gt;http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=7721</a>
&gt;
&gt; I can not recover, what happened in which timeline afterwards. When I
&gt; recognized, what had happened, the firewall (on the notebook) had crashed,
&gt; boinc-cc had crashed and the box had downloaded several WUs and crashed them.
&gt;
&gt;
&gt; Fortunately, the server stopped the box with the message "No work, Daily Quota
&gt; exceeded".
&gt;
&gt; I have made a copy of the whole BOINC_directory, so, if you need any
&gt; information, any file sent, let me know
&gt;
&gt;
&gt; <a> href="http://www.boinc.dk/index.php?page=user_statistics&amp;project=cpdn&amp;userid=34"&gt;<img> src="http://34.cpdn.sig.boinc.dk?142"&gt;</a><img> src="http://75.pah.sig.boinc.dk?142"&gt;<img> src="http://4238.sah.sig.boinc.dk?142"&gt;
&gt;

Honza,

the room temperatur has been around 27 ° Celsius.

About the real cpu-temperatur I can't say anything, the board seems to have no feedback-function to bios (or bios doesn't show it.

The fan was running high (Level 3 of 4), but thats normal, when I'm running Predictor or Seti

When I run Predictor / Seti or CPDN for quit a while and then suspend boinc, it only takes 3 minutes, than the fan switches off. So, I think, the CPU can't be very hot, because then the fan would run longer and with more speed.

My hope is, that it is a real -5, as Carl has mentioned, an unstable modell, not related to CPU-Problem

<a href="http://www.boinc.dk/index.php?page=user_statistics&amp;project=cpdn&amp;userid=34"><img src="http://34.cpdn.sig.boinc.dk?142"></a><img src="http://75.pah.sig.boinc.dk?142"><img src="http://4238.sah.sig.boinc.dk?142">

&gt; When I run Predictor / Seti or CPDN for quit a while and then suspend boinc,
&gt; it only takes 3 minutes, than the fan switches off. So, I think, the CPU can't
&gt; be very hot, because then the fan would run longer and with more speed.

If it takes as long as 3 minutes for your system to return to idle temps, your actual load temps could be quite high.

<a><img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=18"></a>