Comment # 3
on bug 6027
from [email protected]
Even more important (I forgot about this when I added the previous comment)
Some SMB2 messages are "special", and should _never_ play a role in "Service
Response Time" (SRT) measurements. That includes: (1) All SMB2 cancel
messages, and (2) All SMB2 oplock_break messages.
1: SMB2 cancel is sent from client to server after some client request and
(normally) before the server response, to ask the server to finish up the
request, possibly aborting it's operations if necessary. The original request
still gets a response (albeit sooner) and typically with STATUS_CANCELLED.
2: SMB2 oplock_break messages are "reverse" messages, starting with the server
sending and oplock_break "request" to the client asking to recall some cache
delegation. The client may respond by sending an oplock_break "ACK", or if the
client no longer needs the open file, it may instead close the handle.
Given their role in the protocol, it makes no sense for either of the above
message types to take any part in computation of service response times.
I'm not sure how or where SRT computation takes place, or I might offer to fix
this.
You are receiving this mail because:
- You are watching all bug changes.
