[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: Syncrepl: full sync vs. delta



Howard Chu wrote:
I believe neither of these are sufficient.  One cannot
predict which of multiple modifications commits first
(with or without retries).  Which is why the lowest
outstanding CSN is to be used for the cookie...
We could brute force it and fully serialize writes...

As it turns out, all writes to the log database are already serialized. (Makes sense, since it's a flat structure and we're effectively just appending new entries onto the end.) As such, we don't lose anything (more) by serializing up front.

In fact, running test008 with an accesslog overlay that uses a single mutex on the write ops actually completes faster than normal, because back-bdb never hits any deadlocks. (Quite a bit faster, 2x at least.) So I guess this approach is acceptable.

--
 -- 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/