3709721 - OOM runtime dump triggered even when the allocated memory is less than process allocation limit

• Out of memory (OOM) dumps were triggered even when the allocated memory was less than the process allocation limit (PAL).
• High heap memory fragmentation was observed, exceeding acceptable levels.
• In the OOM rtedumps, the sum of all the process allocation limit is higher than the global allocation limit. Example:
  
  GLOBAL_ALLOCATION_LIMIT (GAL) = 239.65gb (257324326912b), SHARED_MEMORY = 53.52gb (57469972480b), CODE_SIZE = 3.13gb (3369938944b), NVM_ALLOCATED_SIZE = 32.25gb (34637103104b), OOM_RESERVATION_ALLOCATOR_SIZE = 96.14mb (100810752b) 
  
  PID=3917966 (hdbnameserver), PAL=146.48gb (157286400000b), AB=24.78gb (26615037952b), UA=0, U=2.93gb (3147138916b), FRO=901.35mb (945139712b), FSL=2.62gb (2818572288b), FMC=0b
  PID=3918216 (hdbcompileserver), PAL=146.48gb (157286400000b), AB=4.19gb (4507324416b), UA=0, U=207.32mb (217391321b), FRO=0b, FSL=0b, FMC=0b, Blocked=1.14gb (1226760192b)
  PID=3918219 (hdbpreprocessor), PAL=146.48gb (157286400000b), AB=4.21gb (4524916736b), UA=0, U=204.06mb (213978866b), FRO=0b, FSL=0b, FMC=0b
  PID=3918266 (hdbindexserver), PAL=146.48gb (157286400000b), AB=56.22gb (60369088512b), UA=0, U=52.34gb (56204055241b), FRO=29.08gb (31226462208b), FSL=0b, FMC=0b
  PID=3918263 (hdbindexserver), PAL=48.82gb (52428800000b), AB=30.70gb (32973443072b), UA=0, U=4.85gb (5216911111b), FRO=2.33gb (2512138240b), FSL=896mb (939524096b), FMC=0b
  PID=3920059 (hdbwebdispatcher), PAL=146.48gb (157286400000b), AB=4.46gb (4797825024b), UA=0, U=455.25mb (477364544b), FRO=0b, FSL=0b, FMC=0b
  PID=3920052 (hdbdiserver), PAL=146.48gb (157286400000b), AB=4.20gb (4510011392b), UA=0, U=224.55mb (235459496b), FRO=0b, FSL=0b, FMC=0b 
  
  Sum of Allocated memory (AB), Shared Memory, and NVM allocated size = 214.53 GB 
  
  In this case, the indexserver with PAL = 146.48 GB could only allocate 56 GB due to available memory already allocated among other services.
  Since the process allocation limit for this indexserver had not been reached, the system tried to allocate 28 GB for some memory intensive operations which resulted in allocated memory going above GAL resulting in OOM.

SAP HANA 2.0

memory issue, heap memory fragmentation, global allocation limit, PAL, GAL, multitenant database container, MDC, out of memory, OOM dump, SAP HANA, memory optimization, defragmentation, database restart. , KBA , HAN-DB , SAP HANA Database , Problem