Re: Extended checkpoint times

Michael Segel wrote: > > anonymous wrote: > > > Hi, > > [SNIP] > > From the informix manual it states checkpoints wait for critical code > > sections to complete before it can complete > > so in theory if an application hung for whatever reason inside a > > critical code section it could hold up a checkpoint > > - forever? or until informix decided the application was not responding? Could be forever. GIGO is as GIGO does. Actually, critical sections are few, far between, and shortlived. They have nothing to do with database/table/row locks but with the acquisition of server resources which are controlled by semaphores known as latches. So, if one has a semaphore latch on an LRU so that the process can acquire a free buffer from that LRU the latch is released once the buffer is moved from the head of the LRU's clean buffer list to either the end of that list, if it is to be used for a page read only, or to the end of the LRU's dirty buffer list if it is to be written/overwritten. This is a momentary thing. You can check latch status with onstat -s. Art S. Kagel