Re: HELP I'm in crisis

In article <329852CE.4ECD@netcomuk.co.uk>, Neil Truby <ntruby@netcomuk.co.uk> writes >Collalti, Fernando wrote: >> >> Hello, >> >> System: HP HP9000 E35, 132MB RAM, Online 7.12.UC3 >> These are my NEW Shared Memory Parameters: >> LOCKS 2000 # Maximum number of locks >> BUFFERS 30000(old=5000) # Maximum number of shared buffers I've configured a system with 128 users and 700Mb of RAM and only used BUFFERS=15000! Sounds like BUFFERS is set way too high. >> NUMAIOVPS # Number of IO vps >> PHYSBUFF 32 # Physical log buffer size (Kbytes) >> LOGBUFF 32 # Logical log buffer size (Kbytes)>> LOGSMAX 6 # Maximum number of logical log files >> CLEANERS 6 # Number of buffer cleaner processes >> SHMBASE 0x0 # Shared memory base address >> SHMVIRTSIZE 60000(old=8000) # initial virtual shared memory segment size >> SHMADD 8192 # Size of new shared memory segments What does onstat -g seg show when the system 'sits'? >> (Kbytes) >> SHMTOTAL 0 # Total shared memory (Kbytes). 0=>unlimited >> CKPTINTVL 300 # Check point interval (in sec) >> LRUS 12 # Number of LRU queues Shounds rather large - how many CPU's do you have? >> LRU_MAX_DIRTY 70 # LRU percent dirty begin cleaning limit >> LRU_MIN_DIRTY 60 # LRU percent dirty end cleaning limit Set these to 40 and 30. >> LTXHWM 50 # Long transaction high water mark>> percentage >> LTXEHWM 60 # Long transaction high water mark >> (exclusive) >> TXTIMEOUT 0x12c # Transaction timeout (in sec) >> STACKSIZE 32 # Stack size (Kbytes)>> >> I've migrated from Informix-SE 5.04 to Informix-OnLine 7.12.UC3 >> >> With SE, the system was slow, but I never had problems. Now >> After 20-30 users are using the system(with the same client/server >> application) OnLine adds new shared memory segment (size 8388608) >> continuosly!!! >> >> After some hours, the system "sits" down and I need to take OnLine >> OFF Line. I'm taking the OnLine 7.12 "on-line"with the NEW above values. >> I've change BUFFERS from 5000 to 30000 and SHMVISTSIZE from 8000 >> to 60000. It seems that the application that was running OK on >> SE, now does not release memory until it doesn't disconnetcs >> from the Database engine....Why???? Are my new parameters ok? Correct - it is a performance enhancement. Online assumes that if the session previously needed say 2MB that it may well need this amount again. Rather than wasting time continually adding and freeing the memory it waits until the session disconnects and then frees all the memory when it is sure it will not be needed again. Try running onmode -F under cron every 15 minutes. This will free any unused memory when it runs. >> I've got 30 users complaining !!!!! HELP!!!! >> >> Please help. I'm doing this installation away from my office. >> >> fernando.collalti@orbit.net > >I haven't got my manuals with me at home so I can't help much. But when >modifying your BUFFERS you should have taken the opportunity to >dramatically reduce your LRU_MAX_DIRTY and LRU_MIN_DIRTY values. If Agreed. >you're concerened to limit the number of segments added, cap the total >available with the SHMTOTAL parameter. > Sorry :-> but I had never agreed with setting SHMTOTAL to anything but 0. If you do and something needs the memory than it gets an error (E.g. a SELECT gets a message "No memory for query processing". and fails. I would rather have the operation complete and run slowly due to swapping then fail completly. >Why does the system "sit"? Any clues from onstat -u, onstat -k etc? > onstat -b, onstat -C, onstat -D, onstat -F, onstat -k, onstat -l, onstat -p onstat -R, onstat -u all these would help. Do not send the output from onstat -a - with thousands of buffers it will be very long!!. >Neil Truby >Chase Bank >Bournemouth, UK -- David Williams