How do you avoid getting the LINCS warning on MD run while...

Hi,I have a big problem with my protein. For studying the role of one residue in enzyme activity, I have to apply dihedral restraints.Therefore, I did all steps and performed the restraints for minimizing energy with emstep =0.01, equilibrating with dt = 0.001 but when I went through the production MD and run with dt = 0.001 after 328 ps I got LINCS warning with this error: Bond length not finite.I checked every step and I am sure my system is converged and equilibriated (with using g_energy command) and read mailing list but still I do not know how to fix it.I changed dt to 0.0005 but when I run this job on cluster it takes much longer than usual. I'll be very thankful if you show me what could be wrong with my initial system.Can anyone help me with this problem?I really appreciate it.Here is my .mdp file:define = -DDISREintegrator = mdtinit = 0dt = 0.0005nsteps = 5000000comm_mode = Linearnstcomm = 1comm_grps = Protein Non-Proteinnstxout = 0nstvout = 250000nstfout = 250000nstlog = 100nstenergy = 500energygrps = Systemnstlist = 1ns_type = gridpbc = xyzrlist = 1coulombtype = PMErcoulomb = 1vdwtype = cut-offrvdw = 1DispCorr = EnerPresfourierspacing = 0.12pme_order = 4ewald_rtol = 1e-5tcoupl = nose-hoovertc_grps = Protein Non-Proteintau_t = 0.1 0.1ref_t = 310 310pcoupl = parrinello-rahmanpcoupltype = isotropictau_p = 2.0compressibility = 4.5e-5ref_p = 1.0gen_vel = noconstraints = all-bondsconstraint_algorithm = LINCScontinuation = yeslincs_order = 4lincs_iter = 1lincs_warnangle = 30;dihedral restraintsdihre = yesdihre_fc = 100dihre_tau = 0.0nstdihreout = 50