On 04/07/2011 02:40 PM, Mark Goodwin wrote:
I'd like to propose another pcp-3 release (say 3.6.0 or
maybe 3.5.1) to (a) pull in Ken's event record instance support,
and (b) to add a few additional metrics that were mentioned as
missing in Frank's comparision of Sigar Vrs PCP.
(a) AFAICT is currently only in Ken's pcp4 branch, specifically
the following commits (is this all Ken, other than QA?) :
e44575b312c3f464bab03c8c9210b0a48ab0e243
ffe97dd82334686ca71099f8e1ba7befd2a2ef18
fbb0350513343361f8816e6ff9b99bd1ff07e0b8
fa60b6d594538294285a360fb1ebd277cbd74a2a
Hmm, commit fbb0350513343361f8816e6ff9b99bd1ff07e0b8 involves
an API change (which is probably why it's in you pcp4 branch Ken) :
/* unpack a PM_TYPE_EVENT value into a set on pmResults */
-extern int pmUnpackEventRecords(pmValueSet *, pmResult ***);
+extern int pmUnpackEventRecords(pmValueSet *, int, pmResult ***);
Is anyone using this yet (David?). Any creative thoughts to work
around it? Maybe we could keep this for the singular instance case:
pmUnpackEventRecords(pmValueSet *, pmResult ***)
and add a new API function for event metrics with an indom :
pmUnpackEventRecordsInstance(pmValueSet *, int, pmResult ***)
where pmUnpackEventRecords(v, r) would just call
pmUnpackEventRecordsInstance(v, 0, r).
Cheers
|