Thanks for clarifying Ken.
On 11/18/2015 02:32 PM, Ken McDonell
wrote:
See the pmlogextract code to see how to generate an appropriate
timestamp for the <mark> record.
I have a couple of observations and questions:
- It looks like the timestamp for the <mark> record is 1
millisecond after that of the previous record in the stream. I
know that this is extremely unlikely, but I suppose that it's
possible that, in a multi-archive context, the next archive
begins less than 1 millisecond after the previous one ends.
Should we worry about that and perhaps interpolate the tmestamp
of the <mark> record in that case? It also seems possible
that within pmLogExtract, the next record might occur less than
a millisecond after the previous one. Is this a problem?
- From what I can tell, it looks to me like pmLogExtract could
generate more than one l<mark> record in the same gap.
This would be the (unlikely?) case where more than one archive
ends at almost the same time and all end within the gap between
the previous record and the next one. Now since we're not
planning on allowing overlapping archives for multi-archive
contexts, this will never happen there. I'm just interested in
whether I'm understanding pmLogExtract correctly and whether
this was known and/or intended.
Thanks,
Dave
|
|