Repository navigation
ANAS 0.4.5
ANAS 0.4.5
A bug release: Hybrid RAID scrubs on real pools, and two user and group
fixes reported on 0.4.4.
Hybrid RAID scrubs finish again
Since 0.4.4 the periodic Hybrid RAID scrub failed on every pool whose
checksum pass takes longer than five minutes, which is every pool on
spinning disks: phase 2 (the btrfs checksum scrub) was killed by the
five-minute command deadline 0.4.4 introduced for the daemon's commands,
and the job ended with a generic "suspended pool or hung disk" sentence
(GitHub #81). The checksum pass now runs with no deadline, as the long
operation it is, and the scrub reports its progress while it runs.
If a scrub failed this way on 0.4.4, the kernel left the pool's scrub
status at interrupted; nothing on the pool was changed. Run Scrub again
on this release, or btrfs scrub start <mountpoint> by hand meanwhile. A
scrub started by hand must finish before Scrub is run here: btrfs allows
one scrub per filesystem at a time, and the screen shows it running.
The Scrubs screen says when a scrub failed or is running
- A failed Hybrid RAID scrub showed exactly like a pool never scrubbed
(GitHub #82). The row now reads "last scrub failed" with the date in
amber, and the reason is the tooltip. A later clean scrub replaces it. - The checksum pass, and a
btrfs scrubstarted by hand, read as running
with the percentage and which pass is running. Previously only the
parity phase showed. - A failed scrub sends one Proxmox notification ("AHR scrub failed")
naming the pool and the reason, through the notification targets
configured on the node.
Users and groups (GitHub #78)
- The default group
users(gid 100), the primary group of every user
ANAS creates, is listed in Share Users and can be chosen in the share
and ACL pickers, as a primary group and as an additional group. Its
membership can be edited; it cannot be deleted. - An account is a system account when its id is outside the share range
or it is a Proxmox VE login, nothing else. A group created through ANAS
and namedadmin,wheelor the like no longer freezes its members,
and a user listed insudoor another system group is still an
ordinary share user. What file access such a membership grants is your
configuration, as before. A hand-made admin login appears in Share
Users and can be disabled there (reversible from the node's shell as
root); deleting it asks for the confirm code like any user.
Documentation
A new page, docs/PBS-TASK-LOGS.md, explains why a Proxmox Backup Server
can refuse new backups with an out-of-space error on its root disk while
the datastore has room (the client's per-chunk task logs), how to check,
and how to clean up. ANAS runs nothing on the backup server; the page
also ships an unsupported prune script for operators who want one.
Install
tar xzf anas-0.4.5.tar.gz && cd anas-0.4.5 && ./install.sh
Re-run over an existing install to upgrade. The installer reloads the
Proxmox web server when its hook changed. Hard-reload the browser
afterwards.