zil: add stats for commit failure/fallback

[robn]

[Sponsors: Klara, Inc., Wasabi Technology, Inc.]

Motivation and Context

There's no good way to tell when a ZIL commit fails and falls back to a transaction sync, other than perhaps a throughput drop. This adds counters so we can see when it happens and why.

Description

Adds three new counters to the zil kstat for each of the three places where a ZIL commit can fail and fall back to a full sync. From the comments:

• error: ZIL IO (write/flush) returned an error (see zil_commit_impl())
• stall: LWB block allocation failed, ZIL chain abandoned (see zil_commit_writer_stall())
• suspend: Commit was requested but ZIL suspended (see zil_commit())

# grep ^zil_commit /proc/spl/kstat/zfs/zil
zil_commit_count                4    0
zil_commit_writer_count         4    0
zil_commit_error_count          4    0
zil_commit_stall_count          4    0
zil_commit_suspend_count         4    0

How Has This Been Tested?

Manual. These stats are also used to implement a test case for a PR to change flush behaviour that I'm about to post.

Full ZTS run completed successfully.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Library ABI change (libzfs, libzfs_core, libnvpair, libuutil and libzfsbootenv)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the OpenZFS code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• I have run the ZFS Test Suite with this change applied.
• All commit messages are properly formatted and contain Signed-off-by.

[amotin]

It looks OK. I am not sure if I like "wait" separate from "suspend", since it is just a different way to detect the same.

You could add these stats into zilstat for completeness.

[robn]

It looks OK. I am not sure if I like "wait" separate from "suspend", since it is just a different way to detect the same.

Ahh, good point, I'd missed that subtlety. They're not both suspend.

You could add these stats into zilstat for completeness.

I always forget this exists! Added.