G
George Potemkin
Guest
The -omsize value was set higher than the number of the objects in database (-omsize 2000 vs 1190 objects). OM latch controls Object Manager cache: “ This is the Object to Area mapping information used to show what Object resides in what Area. The Area also anchors the Index Delete Chain ”. Richard T. Banville wrote: It's not just schema changes that can page something into the secondary cache. Any update where something is not found in the primary cache will cause a latch on the secondary cache. It's just that creating a new table forces it to be in the secondary cache. More about OM latch: community.progress.com/.../49024 Problem: after enabling CDC policies we got the high numbers of the OM latch locks and its naps: ~500,000 locks per sec, ~4,000-6,000 naps perc. OM latch was almost 100% busy (the most snapshots in promon shown a latch owner). In fact the OM latch locks were just 2-6 times less than BHT latch locks on any time intervals. In other words there is a correlation between the activities of these two latches. IMHO, it’s very strange - OM latch deals with db objects while BHT latch deals with the activity of db objects. Any explanation? BHT/OM latch lock ration gives a hint of index tree levels.
Continue reading...
Continue reading...