virtual shared memory segment

hello, this is our output from onstat -g seg : Segment Summary: id key addr size ovhd class blkused blkfree 7680 1381386241 80000000 592850944 245712 R* 144714 25 (shared) 1381386241 a3563000 201334784 6760 V 49154 0 46597 1381386242 c2101000 192512 624 M 47 0 15366 1381386243 c2c0a000 33554432 1640 V 8192 0 519 1381386244 c4d30000 33554432 1640 V 8187 5 2056 1381386245 c6d30000 33554432 1640 V 7122 1070 9 1381386246 c8d30000 33554432 1640 V 732 7460 10 1381386247 cad30000 33554432 1640 V 18 8174 11 1381386248 ccd30000 33554432 1640 V 21 8171 12 1381386249 ced30000 33554432 1640 V 2077 6115 79885 1381386250 d0dde000 33554432 1640 V 6 8186 15 1381386251 d2dde000 33554432 1640 V 8 8184 16 1381386252 d4dde000 33554432 1640 V 26 8166 526 1381386253 d6dde000 33554432 1640 V 28 8164 17 1381386254 d8dde000 33554432 1640 V 9 8183 10771 1381386255 dadde000 33554432 1640 V 2 8190 2072 1381386260 e4dde000 33554432 1640 V 2 8190 1561 1381386261 e6dde000 33554432 1640 V 3 8189 Total: - - 1297694720 - - 220348 96472 Our current ONCONFIG has these settings: SHMVIRTSIZE 196608 # initial virtual shared memorysegment size SHMADD 32768 We are being advised to change these settings to this: SHMVIRTSIZE 524000 # now the original size plus about tentimes the original SHMADD SHMADD 192768 # about six times the original SHMADD because we have "15 dynamically-added virtual segments. Ideally, you want only one or two at the most". My question is: is this not a bit high ? (the new settings) Has anyone run into this situation before? This all came about because our checkpoints were taking a LONG time (up to 150 seconds) and the comment to us was "Your checkpoints are long because there is very little processing going on during the checkpoints. In a 2 -sec interval, you are seeing a couple hundred buffers cleaned max". We are currently seeing very fast checkpoints because we run onmode -B and onmode -c on a cron job every 4 minutes.. Thanks in advance! Tom

On 19 Jun 2003 08:27:04 -0700, schmotom@yahoo.com (tomL) wrote: Are you HP? If so, this is probably a major drain on your system. HP memory management doesn't like that many shared memory segments. May not be high depending on what is happening out in userland . . . >hello, >this is our output from onstat -g seg : > >Segment Summary: >id key addr size ovhd class blkused >blkfree >7680 1381386241 80000000 592850944 245712 R* 144714 25 >(shared) 1381386241 a3563000 201334784 6760 V 49154 0 >46597 1381386242 c2101000 192512 624 M 47 0 >15366 1381386243 c2c0a000 33554432 1640 V 8192 0 >519 1381386244 c4d30000 33554432 1640 V 8187 5 >2056 1381386245 c6d30000 33554432 1640 V 7122 1070 >9 1381386246 c8d30000 33554432 1640 V 732 7460 >10 1381386247 cad30000 33554432 1640 V 18 8174 >11 1381386248 ccd30000 33554432 1640 V 21 8171 >12 1381386249 ced30000 33554432 1640 V 2077 6115 >79885 1381386250 d0dde000 33554432 1640 V 6 8186 >15 1381386251 d2dde000 33554432 1640 V 8 8184 >16 1381386252 d4dde000 33554432 1640 V 26 8166 >526 1381386253 d6dde000 33554432 1640 V 28 8164 >17 1381386254 d8dde000 33554432 1640 V 9 8183 >10771 1381386255 dadde000 33554432 1640 V 2 8190 >2072 1381386260 e4dde000 33554432 1640 V 2 8190 >1561 1381386261 e6dde000 33554432 1640 V 3 8189 >Total: - - 1297694720 - - 220348 96472 > >Our current ONCONFIG has these settings: >SHMVIRTSIZE 196608 # initial virtual shared memory>segment >size >SHMADD 32768> >We are being advised to change these settings to this: >SHMVIRTSIZE 524000 # now the original size plus about ten>times the original SHMADD >SHMADD 192768 # about six times the original >SHMADD> >because we have "15 dynamically-added virtual segments. Ideally, you >want only one or two at the most". > >My question is: is this not a bit high ? (the new settings) >Has anyone run into this situation before? > >This all came about because our checkpoints were taking a LONG time >(up to 150 seconds) and the comment to us was "Your checkpoints are >long because there is very little processing going on during the >checkpoints. In a 2 -sec interval, you are seeing a couple hundred >buffers cleaned max". > >We are currently seeing very fast checkpoints because we run onmode -B >and onmode -c on a cron job every 4 minutes.. > >Thanks in advance! >Tom

John Carlson wrote: > On 19 Jun 2003 08:27:04 -0700, schmotom@yahoo.com (tomL) wrote: > > Are you HP? If so, this is probably a major drain on your system. HP > memory management doesn't like that many shared memory segments. > > > May not be high depending on what is happening out in userland . . . > > > >>hello, >>this is our output from onstat -g seg : >> >>Segment Summary: >>id key addr size ovhd class blkused >>blkfree >>7680 1381386241 80000000 592850944 245712 R* 144714 25 >>(shared) 1381386241 a3563000 201334784 6760 V 49154 0 >>46597 1381386242 c2101000 192512 624 M 47 0 >>15366 1381386243 c2c0a000 33554432 1640 V 8192 0 >>519 1381386244 c4d30000 33554432 1640 V 8187 5 >>2056 1381386245 c6d30000 33554432 1640 V 7122 1070 >>9 1381386246 c8d30000 33554432 1640 V 732 7460 >>10 1381386247 cad30000 33554432 1640 V 18 8174 >>11 1381386248 ccd30000 33554432 1640 V 21 8171 >>12 1381386249 ced30000 33554432 1640 V 2077 6115 >>79885 1381386250 d0dde000 33554432 1640 V 6 8186 >>15 1381386251 d2dde000 33554432 1640 V 8 8184 >>16 1381386252 d4dde000 33554432 1640 V 26 8166 >>526 1381386253 d6dde000 33554432 1640 V 28 8164 >>17 1381386254 d8dde000 33554432 1640 V 9 8183 >>10771 1381386255 dadde000 33554432 1640 V 2 8190 >>2072 1381386260 e4dde000 33554432 1640 V 2 8190 >>1561 1381386261 e6dde000 33554432 1640 V 3 8189 >>Total: - - 1297694720 - - 220348 96472 >> >>Our current ONCONFIG has these settings: >>SHMVIRTSIZE 196608 # initial virtual shared memory>>segment >>size >>SHMADD 32768>> >>We are being advised to change these settings to this: >>SHMVIRTSIZE 524000 # now the original size plus about ten>>times the original SHMADD >>SHMADD 192768 # about six times the original >>SHMADD>> >>because we have "15 dynamically-added virtual segments. Ideally, you >>want only one or two at the most". >> >>My question is: is this not a bit high ? (the new settings) >>Has anyone run into this situation before? >> >>This all came about because our checkpoints were taking a LONG time >>(up to 150 seconds) and the comment to us was "Your checkpoints are >>long because there is very little processing going on during the >>checkpoints. In a 2 -sec interval, you are seeing a couple hundred >>buffers cleaned max". >> >>We are currently seeing very fast checkpoints because we run onmode -B >>and onmode -c on a cron job every 4 minutes.. >> >>Thanks in advance! >>Tom > > You are already using the virtual memory, so to have it in one segment is just ... "better". Having gone through the figures, I think someone can't add up (could be me) : SHMVIRTSIZE of 196608k plus 15 * SHMADDS of 32768 I make that 688128! So I would suggest setting SHMVIRTSIZE to 756 Mb SHMVIRTSIZE 774144 Whether you increase SHMADD or not isn't the point really, you want to try and stay within one virtual segment. I would probably leave SHMADD ad 32Mb and keep an eye on whether you do allocate any extra segments, and increase SHMVIRTSIZE by the number of segments added. Mash it up.