Originally posted by: ShaCplex

Hi Alex/Daniel,

My QP model (Cplex-Java) sometimes hangs or getting stuck after the last line;

 last part of the engine log is shared below;

 Number of nonzeros in lower triangle of Q = 768874
Using Approximate Minimum Degree ordering
Total time for automatic ordering = 0.42 sec. (100.76 ticks)
Summary statistics for factor of Q:
   Rows in Factor            = 2207524
  Integer space required    = 2207524
 Total non-zeros in factor = 2977178
 Total FP ops to factor    = 4518046
Tried aggregator 1 time.
QP Presolve eliminated 825246 rows and 822737 columns.
Aggregator did 2199712 substitutions.
 Reduced QP has 1500248 rows, 3142556 columns, and 8775659 nonzeros.
 Reduced QP objective Q matrix has 1431614 nonzeros.
 Presolve time = 18.92 sec. (3869.35 ticks)
 Parallel mode: using up to 6 threads for barrier.
Number of nonzeros in lower triangle of A*A' = 10271639
 Using Approximate Minimum Degree ordering
Total time for automatic ordering = 26.40 sec. (19416.68 ticks)
Summary statistics for Cholesky factor:
   Threads                   = 6
Rows in Factor            = 1500248
 Integer space required    = 10646719
Total non-zeros in factor = 331908873
 Total FP ops to factor    = 4008117166877

after this line, its not proceeding and doesn't provide the solution. we then have to restart our CPLEX server. Could you please share your thoughts on this issue?

Thanks,

Shahul

Originally posted by: ShaCplex

Hi,

Can you guys help me on this?

Regards,

Shahul

What is the CPU usage? Does CPLEX still consume CPU time or does it look completely stuck? Does the problem persist if you switch to a single thread?

Can you share one of the offending model instances?

Originally posted by: ShaCplex

Thanks Daniel. will check and get back to you on this.

At the outset, it looks CPU Utilization reaches even up to 99.89% (CPU Utilization: %' (value= 99.89) on node).

would this be a potential cause?

Thanks,

Shahul

It could be that CPLEX is stuck in a loop. Do you have any indication of numerical challenges for your model? Again, can you share a model instance that shows this behavior?