On Tue, 7 Jan 2014 20:39:56 -0500, Frank Ch Eigler wrote:
fche> 1.1) pmlogger needs to learn to write its output with what IIRC kenj
fche> has referred to as "semantic units", ie., proper use sequencing of
fche> write(2), fdatasync(), to put interdependent data on disk correctly.
This was the biggest bugbear of NAS Manager (for those who remember)
which tried to provide historical and live data using archives as main
source of information. I'd imagine this could be solved given more
then .33 of engineer available to work on PCP. Still, the nightmares
are still with me and I wasn't the .33 at the time.
fche> 2.1) A new server needs to be written, which would monitor some local
fche> archive files, and serve an extended pcp wire protocol for it (one
fche> that includes archive-like pmSetMode operations). It would advertise
fche> the pcp hostname (or pmmgr-style hostid) to the network, so clients
fche> can find the right host data (probably one tcp port per archive, or
fche> else multiplexed over a single tcp port and identifying the host
fche> during startup negotiation). The clients could keep using
fche> PM_CONTEXT_HOST but permit PM_MODE_BACK etc. to forward -S/-T times -
fche> ie. no requirement for a new PM_CONTEXT_UNIFIED.
What would happen if pcp client is using PM_CONTEXT_HOST with new
protocol? Will it be banned or will it be silently dropped back to
current "live" sematic?
max
|