Checkpoint Issues in 10.00.FC8

We have an issue with periodic long checkpoint on a server running 10.00.FC8 (AIX 5.3.0.0). We see that the buffers are drained, nothing is in a critical state, and the number of dirty buffers will go to zero, but the checkpoint remains blocked for several more seconds. From the onconfig, here's our BUFFERPOOL statement: BUFFERPOOL size=4K,buffers=650000,lrus=48,lru_min_dirty=0.000000,lru_max_dirty=0.500000 Is there something going on in the engine that is causing the checkpoint to remain blocked, even though the number of dirty buffers has dropped to zero? Thanks, [cid:image003.jpg@01CA8DF4.120CB600]Jeffrey J. Mitchell Database Administrator, West Interactive Corporation 11650 Miracle Hills Drive, Omaha NE 68154 402-716-0500 | Cell 402-321-7443 | jjmitchell@west.com<mailto:jjmitchell@west.com> This electronic message transmission, including any attachments, contains information from West Corporation which may be confidential or privileged. The information is intended to be for the use of the individual or entity named above. If you are not the intended recipient, be aware that any disclosure, copying, distribution or use of the contents of this information is prohibited. If you have received this electronic transmission in error, please notify the sender immediately by a "reply to sender only" message and destroy all electronic and hard copies of the communication, including attachments.

You wrote: >We have an issue with periodic long checkpoint on a server running 10.00.FC8 (AIX 5.3.0.0). >We see that the buffers are drained, nothing is in a critical state, and the >number of dirty buffers will go to zero, but the checkpoint remains blocked for >several more seconds. >From the onconfig, here's our BUFFERPOOL statement: BUFFERPOOL size=4K,buffers=650000,lrus=48,lru_min_dirty=0.000000,lru_max_dirty=0.500000 >Is there something going on in the engine that is causing the checkpoint to >remain blocked, even though the number of dirty buffers has dropped to zero? The simple answer to that question is yes, there is other things the engine does in the checkpoint code after it does the bufferpool flush and prior to it clearing the checkpoint blocked state. Do you have FAST_RESTART_CKPT_FUZZYLOG set in your onconfig? That in particular appears to potentially cause some additional IO that needs to be done to the physical log prior to freeing up the checkpoint. Unfortunately, prior to version 11, there isn't a lot of additional information about what the main_loop thread is doing while the checkpoint is happening (also version 11 checkpoints no longer block while the buffer pool is flushed due to the new interval checkpoints). You could try to collect several onstat -g stk <thread if for main_loop> outputs after you notice buffers at 0 and the checkpoint hasn't released. However that info won't be reliable if the thread is running. If main_loop thread is running, you would be better off collecting a stack trace with something like pstack. Jacques Renaut IBM Informix Advanced Support ADP team