Handling Memory Limits
Occasionally we run into a situation on development hosts in which the informixserver is running several memory-intensive processes and attempts to procure more memory than the O/S permits. Since we don't set SHMTOTAL (and RESIDENT=0), the informixserver at first fills its tmp dbspaces, then its root dbspace, then just aborts, leaving behind unhelpful assert failure files. If we set SHMTOTAL, however, in the same circumstance the message log is flooded with notices redundantly describing the lack of memory; the informixserver is rendered unusable, not responding even to attempts to kill the offending sessions, and we have to shutdown and restart the instance anyway. Have others experienced this kind of error? Is there any way for the informixserver itself to avoid these extremes in behavior -- seppuku or schizophrenia -- somehow, and ensure that it can handle more adroitly a lack of memory resources? Running IDS 11.50.FC4 on HP-UX 11.23.
On Feb 22, 3:30 pm, "red_val...@yahoo.com" <red_val...@yahoo.com> wrote: > Occasionally we run into a situation on development hosts in which the > informixserver is running several memory-intensive processes and > attempts to procure more memory than the O/S permits. Since we don't > set SHMTOTAL (and RESIDENT=0), the informixserver at first fills its > tmp dbspaces, then its root dbspace, then just aborts, leaving behind > unhelpful assert failure files. If we set SHMTOTAL, however, in the > same circumstance the message log is flooded with notices redundantly > describing the lack of memory; the informixserver is rendered > unusable, not responding even to attempts to kill the offending > sessions, and we have to shutdown and restart the instance anyway. > > Have others experienced this kind of error? Is there any way for the > informixserver itself to avoid these extremes in behavior -- seppuku > or schizophrenia -- somehow, and ensure that it can handle more > adroitly a lack of memory resources? > > Running IDS 11.50.FC4 on HP-UX 11.23. I was hitting memory allocation limits for informix 32 bits and changing SHMBASE ( to 0xB000000L) helped it allocate more memory. The machine notes might have an example for your platform. ex: /home/ informix/release/en_us/0333/ids_machine_notes_11.50.txt But I guess the only real solution would be to upgrade to 64bits
On 22/02/2010 19:47, brenddie wrote: > On Feb 22, 3:30 pm, "red_val...@yahoo.com"<red_val...@yahoo.com> > wrote: >> Occasionally we run into a situation on development hosts in which the >> informixserver is running several memory-intensive processes and >> attempts to procure more memory than the O/S permits. Since we don't >> set SHMTOTAL (and RESIDENT=0), the informixserver at first fills its >> tmp dbspaces, then its root dbspace, then just aborts, leaving behind >> unhelpful assert failure files. If we set SHMTOTAL, however, in the >> same circumstance the message log is flooded with notices redundantly >> describing the lack of memory; the informixserver is rendered >> unusable, not responding even to attempts to kill the offending >> sessions, and we have to shutdown and restart the instance anyway. >> >> Have others experienced this kind of error? Is there any way for the >> informixserver itself to avoid these extremes in behavior -- seppuku >> or schizophrenia -- somehow, and ensure that it can handle more >> adroitly a lack of memory resources? >> >> Running IDS 11.50.FC4 on HP-UX 11.23. > > I was hitting memory allocation limits for informix 32 bits and > changing SHMBASE ( to 0xB000000L) helped it allocate more memory. The > machine notes might have an example for your platform. ex: /home/ > informix/release/en_us/0333/ids_machine_notes_11.50.txt > But I guess the only real solution would be to upgrade to 64bits Hi, So, how much RAM is available? How much memory is the INFORMIXSERVER "normally" running in? "Memory intensive operations" ... big sorts? Loads of PDQ? Archive running? Can you post the top of the very first assert failure? Or at least the online.log where the engine afs?