If you use the STARJOIN facility's IN-SET strategy, SAS Scalable Performance Data Server might disconnect.

Click the Full Code tab in this SAS Note to see an example of code that produces the following error in the SAS SPD Server log:

05/30/16 08:44:05:(10029) SPDSBASE- STARINSET: SETUP FACT JOIN FINISHED
05/30/16 08:44:05:(10029) SPDSBASE- ERROR: Access Violation
05/30/16 08:44:05:(10029) SPDSBASE-
/saswork/SPDS52/SASScalablePerformanceDataServer/5.2/bin/spdsbase() [0x6a1b3e]
/saswork/SPDS52/SASScalablePerformanceDataServer/5.2/bin/spdsbase() [0x69cba9]
/saswork/SPDS52/SASFoundation/9.4/sasexe/tkmk.so(bktExcept+0x62) [0x7f338922df92]
/saswork/SPDS52/SASFoundation/9.4/sasexe/tkmk.so(bkt_signal_handler+0x144) [0x7f338922e5d4]
/lib64/libpthread.so.0() [0x340980f710]

To circumvent this issue, disable STARJOIN optimization in your SAS code by setting the following macro variable:

Click the Hot Fix tab in this note to access the hot fix for this issue.

Operating System and Release Information

%let dom=spds51;
%let host=myhost.domain;
%let portno=5200;

libname foo sasspds "&dom" host="&host" serv="&portno" user='admin' password='secret' IP=YES LIBGEN=YES;

/*table1*/

data foo.fixe_SUPPORTSAS;
finess2='754218071';
ident='0c2fb5';
char2='A';
output;
run;

proc datasets lib=foo;							
modify fixe_SUPPORTSAS;
index create ident;
index create finess2;
quit;


/*table2*/

data foo.um_SUPPORTSAS;
finessGeo2='754218099';
ident='0c2fb5';
rum='01';
output;
run;

proc datasets lib=foo ;							
modify um_SUPPORTSAS;
index create ident;
index create cleru=(ident rum);
quit;

/*table3*/

data foo.acte_SUPPORTSAS;
ident='0c2fb5';
rum='01';
char3='A';
output;
run;

proc datasets lib=foo;							
modify acte_SUPPORTSAS;
index create ident;
index create cleru=(ident rum);
quit;



%let spdssqlr= _method nostarjoin=0 DETAILS="stj$";


proc sql; 

create table toto as select t2.FinessGeo2,
 
case  when SUBSTR(t1.char2, 1, 5) = 'A' 
then 'mastec' else 'autre' end as Type,
 
   COUNT(t2.ident) as COUNT_of_ident 
     from foo.FIXE_SUPPORTSAS  t1, 
          foo.UM_SUPPORTSAS t2, 
          foo.ACTE_SUPPORTSAS  t3
where ((t1.ident = t2.ident) and (t1.ident = t3.ident)) and 
     (t1.finess2 = '754218071') and (( SUBSTR(t3.char3, 1, 1) 
in ('A', 'B', 'C','D') ) and (t3.rum = '01')) group by t2.FinessGeo2;
quit;

A fix for this issue for SAS Scalable Performance Data Server 5.2 is available at: