Re: Estimates for syncprov sessionlog and checkpoint

[Howard Chu <hyc@symas.com>]

Andreas wrote:
> How could one estimate good values for syncprov's sessionlog and
> checkpoint values?
>
> The way I see it, sessionlog should be big enough to store all possible
> entries that are going to be added between replication intervals, is
> that more or less it?
>   

No. Read the slapo-syncprov(5) manpage again. It clearly says "All write 
operations (except Adds) are recorded..." Adds don't need to be 
explicitly saved in the session log; the added entry's entryCSN already 
makes it obvious that the entry is new.

> And what is inside the sessionlog, the operation plus the data I
> presume?
>   

Just the entryUUID, entryCSN, and the type of operation. The actual data 
resulting from the operation is of course present in the entry itself.
> Finally, what about the checkpoint? The admin guide says the sessionlog
> is stored in memory, so what does a checkpoint mean in this context?
>   

The manpage is pretty explicit about that too. "After a write operation 
has succeeded, write the contextCSN to the underlying database..."

--
 -- Howard Chu
 Chief Architect, Symas Corp.  http://www.symas.com
 Director, Highland Sun        http://highlandsun.com/hyc
 OpenLDAP Core Team            http://www.openldap.org/project/