Incremental checkpoint and SCN

Question

Hi,
I am getting messages of incremental checkpoint in my alert logs with some scn.
>
Completed checkpoint up to RBA [0x125de6.2.10], SCN: 445135162445
>
Does this mean that all dirty blocks which have had their initial changes before this SCN(445135162445) will be written to disk so that instance recovery can begin from the SCN from which checkpoint has completed.Or is it the other way like the incremental checkpoint has occured at scn 445135162445.
Sekar

Answer 1

user13485610 wrote:
As per my knowledge, the checkpoint is classified as below (correct me if I am wrong somewhere)
Checkpoint types can be divided as INCREMENTAL and COMPLETE.
Also COMPLETE CHECKPOINT can be divided further into
PARTIAL and FULL.
It would be convenient to have a reference to the documents where you picked up this information. There may be further reading in them that clarifies the meaning. The terms have been around for a long time, of course, but it's always hard to get any sort of definitive description together - in your case, for example, you don't make any comment about which checkpoints lead to high priority writes and which to low, but the description of any type of checkpoint isi incomplete without some reference to the write priority.
As far as classifying checkpoints by name - I'm not too concerned that there is still some confusion in the different way that people name or categorise them, provided that they can describe what's going on to ensure that there is no ambiguity. In this context I think there are only options to consider:
a) does the particular type of checkpoint walk along the checkpoint queue (CKPTQ) in order to pick the blocks that need to be written to disc.
b) does the particular type of checkpoint use a different queue (such as an object queue or file queue) to pick the blocks that need to be written to disc.
c) is there any other mechanism for picking the blocks to be written - such as walking the LRU and identifying all dirty blocks.
To my mind, an incremental checkpoint should probably have a definition that says it walks the checkpoint queue.
I dislike the term "complete" if it then leads to the option for "partial" - how much clarity can you read into the statement "at this point Oracle does a partial complete checkpoint" (or should that be a "complete partial checkpoint") - but I can understand the need for a term of that sort to distinguish a checkpoint that is based on one of the other queues.
But my doubt is mentioned below.
2.At the time of log switch - Sometimes log switches may trigger a complete checkpoint , if the
next log where the log switch is to take place is Active.
Why is this behaves in this fashion? (Any internal thoughts on this please)This, in part, is why I'd like to see the reference document - I think that the term "complete" may have been given a different meaning at this point. If the logfile you want to use is still active checkpoint activity MUST take place urgently, but it need only be a checkpoint that walks the CKPTQ up to the point where the content of the target redo log can be discarded. This is no different from any other checkpointing due to log file switch - but it could have a higher degree of urgency. (The need to differentiate this special case on log file switch probably came about at the time that Oracle stopped triggering an automatic checkpoint at every log file switch.)
Regards
Jonathan Lewis