[LUDOC-142] Watching the Client RPC Stream (Ch 31)

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Labels: QContent
• Business Value: 4.9
• Severity: 3
• Assignee: Linda Bebernes (Inactive)
• Reporter: Brett Lee (Inactive)

Description

Lustre 2.x manual, section:
31.2.2. Watching the Client RPC Stream

The parameter "offset" doesn't have any description.

It should have a description, ideally a quick summary. It should also point to the section in the manual that deals with offsets (currently 31.2.3. Client Read-Write Offset Survey).

Comments

[Linda Bebernes (Inactive)]

Issue has been fixed and merged.

[Linda Bebernes (Inactive)]

Ready for review at http://review.whamcloud.com/#/c/7362

[Linda Bebernes (Inactive)]

"Offset" needs to be defined in the table in Section 31.2.2: Watching the Client RPC Stream.

[Brett Lee (Inactive)]

I misinterpreted.

From Andreas (in a follow up):
It might be that "offset"
> is relative to the previous read/write as you propose, or it could be the offset
> within the 1MB RPC. I would suspect the latter, because this gives a bounded
> range of values, and is the only critical metric for the RPC stack. Doing 1MB
> sized and aligned write RPCs (i.e. "offset = 0" per my definition) but with
> random file offsets should give nearly the same performance vs. 1MB
> sequential writes. In contrast, doing smaller RPCs would cause read-modify-
> write cycles on the underlying RAID device and should be avoided.

Brett is now trying to understand what "bounded range of values" means, and also whether an RPC can be comprised of a several small IO's (each presumably an offset?)...

[Brett Lee (Inactive)]

rpc_stats - "offset" metric refers to reads or writes not accessing the next sequential location, indicating small or random IO.

From Andreas:
this is the offset of the pages within the 1MB "ideal" range, and can either indicate small IOs from the application (whether sequential or not) and/or random IOs for ideally-formed 1MB (or larger) client IOs the offset would always be 0.