best_response client/broker inconsistency

[Olav Syljuasen]

Jobs are not accepted as the broker best_response algorithm yields a zero result. However when the same best_response command is issued on the command-line in the client it consistently yields a non-zero result.

Compare the errors logfile from not accepted job:

<logfile> <log dn="/O=Grid/O=NorduGrid?/OU=nordita.dk/CN=Olav Syljuasen" jid="318431157611862934519903"><time>Wed Sep 6 23:51:28 2006</time><message><Action>SUBMIT_J OB</Action><Count>1</Count><Interval>990000</Interval></message></log>

<log dn="/O=Grid/O=NorduGrid?/OU=nordita.dk/CN=Olav Syljuasen" jid="318431157611862934519903"><time>Wed Sep 6 23:51:28 2006</time><message>tycoon best_resp onse -oUseNetwork=204.123.32,IgnoreHosts?="204.123.32.54 204.123.32.70 204.123.32.50 204.123.32.40 204.123.32.41 204.123.32.73 204.123.32.6204.123.32.35

204.123.32.65

204.123.32.56

204.123.32.73

204.123.32.57

204.123.32.37

204.123.32.42

204.123.32.55

204.123.32.39

204.123.32.64

204.123.32.66

204.123.32.36" 10 990000 1</message></log>

<log dn="/O=Grid/O=NorduGrid?/OU=nordita.dk/CN=Olav Syljuasen" jid="318431157611862934519903"><time>Wed Sep 6 23:51:36 2006</time><message></message></log>

<log dn="/O=Grid/O=NorduGrid?/OU=nordita.dk/CN=Olav Syljuasen" jid="318431157611862934519903"><time>Wed Sep 6 23:51:37 2006</time><message>refunding 10</me ssage></log>

<log dn="/O=Grid/O=NorduGrid?/OU=nordita.dk/CN=Olav Syljuasen" jid="318431157611862934519903"><time>Wed Sep 6 23:51:38 2006</time><message>Error Got no hos ts to fund exiting</message></log>

</logfile>

with the command-line:

nw2[09:03]~/SpinS/Run6904$ tycoon --version tycoon 0.4.1

nw2[08:55]~/SpinS/Run6904$ tycoon best_response -oUseNetwork=204.123.32,IgnoreHosts?="204.123.32.54 204.123.32.70 204.123.32.50 204.123.32.40 204.123.32.41204.123.32.73 204.123.32.6204.123.32.35 204.123.32.65 204.123.32.56 204.123.32.73 204.123.32.57 204.123.32.37 204.123.32.42 204.123.32.55 204.123.32.39 204.123.32.64 204.123.32.66 204.123.32.36" 10 990000 1 204.123.32.81 10 2799.31199939 10

[ThomasSandholm]

this was due to a xen memory allocation problem and has been fixed. the broker has also been rewritten to consume less memory