Cleaners

Please forgive me if this is posted twice ----------------- Online 7.24UC3 Solaris 2.5.1 Given that Cleaners are assigned to chunks in the order that the chunks are created (and I don't know the order) how can I determine which cleaner from onstat -F) is assined to which chunk? I figure from onstat -d, chk 1 is the first chunk created, chk 2 the second and so forth. So cleaner 1 is assinged to chunk 1, cleaner 2 to the second chunk ...etc. But how are cleaners assigned to the mirrored chunks? My root dbspace is mirrored. From the output of onstat -d , I see that the primary chunk is numbered chk/1 and so is the mirrored chunk. Does that mean the first page cleaner is assigned to Both Chunks? I have a situation where a single flusher is slower then molasses during a checkpoint and I need to find out which disk is the culprit. Many Thanks, Steve Faehn

Stephen Faehn wrote: > > Please forgive me if this is posted twice > ----------------- > Online 7.24UC3 > Solaris 2.5.1 > > Given that Cleaners are assigned to chunks in the order that the chunks > are created (and I don't know the order) how can I determine which cleaner > from onstat -F) is assined to which chunk? > > I figure from onstat -d, chk 1 is the first chunk created, chk 2 the second > and so forth. So cleaner 1 is assinged to chunk 1, cleaner 2 to the second > chunk ...etc. > > But how are cleaners assigned to the mirrored chunks? > My root dbspace is mirrored. > From the output of onstat -d , I see that the primary chunk is numbered > chk/1 and so is the mirrored chunk. Does that mean the first page cleaner > is assigned to Both Chunks? > I have a situation where a single flusher is slower then molasses during a > checkpoint and I need to find out which disk is the culprit. > Many Thanks, Steve Faehn Cleaners are assigned not to clean a chunk, during a checkpoint, but to flush all of the dirty pages that the checkpoint thread has gathered from the various LRUS that all belong to a single chunk. As to which cleaner thread is assigned it depends on which one us free at the time a particular chunk is being launched. Better to get some OS level disk/controller monitoring tools and see where the bottleneck is. You can get a rough idea where to look first by clearing stats (onstat -z) shortly before a scheduled checkpoint then run onstat -d and/or onstat -g iof shortly after the checkpoint completes (or if long even while it is running) to see which disks are being hit and which queues are have long queue lengths. If I had to guess I'd look for an older drive with less cache or a slower spindle that the rest. Art S. Kagel