Quick questions for the developers - if one generates a filesystem with
version 2 logs and specifies a stripe unit size, of say 16kB, at this
time, does this reduce the number of transactions that can be in the log
at a given time? Is there a need to increase the log size from the
default to ensure better performance by making sure that the log doesn't
need to be flushed more frequently as a result of this change?
I'm imagining as follows: if a log is, say, 1000 4kB filesystem blocks (as
long picked by a default mkfs command, hypothetically) when using version
1, it can hold 8000 512-byte log entries. The same log in version 2
format presumably holds less entries if everything is done stripe unit
Quite probably I'm just plain confused, having not read how the version
two log code works yet - is there a simple answer?