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

Re: (ITS#3696) Determining server configuration at run-time (RFE)

At 11:10 PM 5/1/2005, ando@sys-net.it wrote:

>> At 10:19 PM 5/1/2005, ando@sys-net.it wrote:
>>>Information about a database being shadow or not could be added to the
>>>database entries.
>> Adding a syncTimeStamp (or shadowTimeStamp) has been on
>> my long todo list for ages.  The presence of this would
>> indicate that the DSE is a shadow copy.
>BTW, do you mean a regular timestamp, or a CSN?


> The former makes sense
>for slurpd, because that's all we have;

Well, the slave could generate a CSN.

>the latter makes sense for
>syncrepl, where the timestamp implies loss in precision.

Well, the slap can produce timestamps at arbitrary precision
(even at finer grained than their clocks... (as there is
no requirement for timestamps to be more accurate than the
clocks allows)).

>We could use a
>monitorShadowTimeStamp for slurpd replication, which essentially echoes
>the latest "now" as used in replog(), and a "monitorSyncReplCSN" for
>syncrepl replication, which essentially uses the si_syncCookie field of
>the syncinfo_t. Or use only the latter, and for slurpd replication turn
>the timestamp into a CSN.

In both cases, the timestamp should be locally generated
(by the slave, not the provider).