Repository navigation
Commit 2fc68f0
committed
test(e2e): record ZFS datasets and LINSTOR volumes in cozyreport
A kubernetes suite can pass and still fail its cleanup because one
node's ZFS pool stays a few hundred MiB short of its baseline. The
report carries the LINSTOR resource list and the pool totals, but
nothing that names the dataset holding the space, so a leftover zvol,
a snapshot and pool metadata cannot be told apart.
Each satellite now lists every dataset of the ZFS pools LINSTOR
records for its node, snapshots included, with the pool's pending
frees. Sizes are exact bytes with avail, since the cleanup check works
in KiB of LINSTOR free capacity and the rounded form loses a sub-GiB
difference on a large pool. The pool names come from LINSTOR rather
than from a name pattern. The controller's volume list is recorded
next to the resource list. Everything runs on every job, so a red
listing can be read against a green one, and goes through the bounded
exec reader like the other LINSTOR reads. Count claims about those
reads, and the time budget in the header, are updated to match.
Signed-off-by: Aleksei Sviridkin <[email protected]>
Assisted-by: LLM1 parent bc400df commit 2fc68f0
3 files changed
Lines changed: 196 additions & 13 deletions
0 commit comments