like we do on a full restore, for that refactor the functionality into
'log_required_tapes' to be reused across the workers
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
Adding an opening line for every group makes parsing the log easier.
We can also remove the 're-sync [...] done' line, because the next
line should be a progress line anyway.
The new output for the sync job/pull logs looks as follows:
- skipped already synced (happens in most jobs, except for first run)
- re-sync of last synced snapshot (if it still exists on source)
- skipped because of transfer-last (if set and skips something)
- sync of new snapshots (if they exist)
Suggested-By: Fabian Grünbichler <f.gruenbichler@proxmox.com>
Signed-off-by: Stefan Hanreich <s.hanreich@proxmox.com>
Specifying the transfer-last parameter limits the amount of backups
that get synced via the pull command/sync job. The parameter specifies
how many of the N latest backups should get pulled/synced. All other
backups will get skipped.
This is particularly useful in situations where the sync target has
less disk space than the source. Syncing all backups from the source
is not possible if there is not enough disk space on the target.
Additionally this can be used for limiting the amount of data
transferred, reducing load on the network.
The newest backup will always get re-synced, regardless of the setting
of the transfer-last parameter.
Signed-off-by: Stefan Hanreich <s.hanreich@proxmox.com>
This reverts commit 6bac67195e and
commit 783e88c2df as they were both
superseded by 569324cb ("provide separate helpers for pub/priv auth
keyring access").
This used to be the case before the switch to the auth api
crate and is required for some helpers where we don't want
to have to setup the complete auth context.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
as otherwise build fails in a clean environment, where no API auth
key is available.
This whole printdoc command injection is quite ugly and causes
headache in general though, we'd be better off if we could do away
with that..
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
The manager, tape and debug admin and debug CLI tools all are meant
to run locally on the PBS directly and often directly by the root
user.
So for convenience they will create a valid API ticket for `root@pam`
to connect via localhost to the API proxy, either transparently if
the UID is 0 (= root) or interactively by asking the password of the
root user.
In d97ff8ae ("use new auth api crate") the API key handling was
reworked, but while the call sites that get the auth keys for signing
a ticket where adapted to use the new auth_keyring helper, the
reworked API keeps the key in shared OnceCell which needs to be
actually setup once per executable to the respective correct key,
i.e., public for the proxy, which only checks signatures, but never
creates them, and private for all those that want to sign tickets;
and that was only added for the prrivileged and public daemon, but
not above mentioned CLI tools.
So all calls into the tools that actually required to create a ticket
for connecting to the local API panicked.
Fix this by adding a call to setup_auth_context in the main entry
point of each of those CLI tools, so that they have the private key
available to be used if the tools is executed by root.
Fixes: d97ff8ae ("use new auth api crate")
Reported-by: Friedrich Weber <f.weber@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
previously the upid would just be used without a file extension when
downloading a task log. this lead to rather strange filenames that
appeared unfamiliar to users as the upid is not very prevalent in the
gui. set a proper file name based on the node name, worker type and a
time stamp instead. also add the ".log" file extension to indicate
that these files contain logs.
Signed-off-by: Stefan Sterz <s.sterz@proxmox.com>
the backup is finished at that point, the only lock clash that is possible when
dropping the exclusive and attempting to obtain a shared lock would be
- the snapshot is pruned/removed
- the backup is in a pre-upgrade process, and the post-upgrade process opens a reader
the first case is OK, if the other invocation wins the race and removes the
snapshot verification is pointless anyway.
the second case means the snapshot is not verified directly after completion
(this fact would be logged in the backup task log), but usable immediately for
pulling/restoring/..
this should decrease the chances of triggering the issues described in #4523
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
Sometimes it can be quite useful to know when exactly a system report
was generated. Adds the following output:
# date -R
Thu, 23 Feb 2023 16:21:12 +0100
Signed-off-by: Lukas Wagner <l.wagner@proxmox.com>
Previously, the same approach as in `proxmox-backup-debug` was used.
With the changes from this commit, realm syncing uses the same method for
waiting for task output as other parts of `proxmox-backup-manager`.
Signed-off-by: Lukas Wagner <l.wagner@proxmox.com>
Previously, if the value of a synced property did not validate properly
(e.g only 1 character in length instead of the required 2), the whole
sync job failed without any useful error message.
In this commit, the values are validated manually by their
respective StringSchema. If the validation fails, the value is
ignored and a warning is displayed in the task log.
In addition to that, some error messages have been improved.
Also, user sync is now more fault-tolerant in general, showing
warnings if something goes wrong while creating/updating a
single user, instead of aborting the whole sync job.
Reported-by: Friedrich Weber <f.weber@proxmox.com>
Signed-off-by: Lukas Wagner <l.wagner@proxmox.com>
Tested-by: Friedrich Weber <f.weber@proxmox.com>
while with rust strings we cannot inject \0, it feels a bit safer to
enforce some basic restrictions, with length and not containing any
slash seems sensible enough.
Admins should not put sensible data as theme-XYZ.css files in
/usr/share (which is normally readable by all system users anyway)
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
adds a theme switcher to the ui and handles the necessary cookies in
the backend.
this requires a bump of the widget toolkit so the necessary widgets
are present.
Signed-off-by: Stefan Sterz <s.sterz@proxmox.com>
the only thing preventing us from using WORM tapes was that we relied
on being able to rewrite the media set label when first using a tape
that was pre-allocated in a media-pool.
so instead of needing to write a meida set label with a special uuid,
just save the pool in the media label itself. This has currently no
downsides, as we're not able to move tapes from one pool to another
anyway.
this makes some checks a bit trickier, as we now have to get the pool
out of the media set label and as a fallback look into the media label.
such new tapes can still be read and restored by older proxmox-bacukp-server
versions. The only thing missing is when a tape labeled with the new
format that has an assigned pool, that pool won't show up when the tape
is inventoried in an old version (but can still be used otherwise).
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
since commit 139acf37 ("tape: inventory: skip unassigned tapes")
we skip unassigned tapes (special all-zero media-set uuid) when we look
for a catalog. We accidentally skipped storing it in the inventory
completely, which means we never inventoried completely empty tapes.
to fix that, simply move the check below the inserting in the inventory
Fixes: 139acf37 ("tape: inventory: skip unassigned tapes")
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
When manually adding a user, `lookup_authenticator` is called
to verify whether a realm actually exists. Thus is is necessary
to have dummy implementation for the `ProxmoxAuthenticator` trait
for OpenID realms.
Signed-off-by: Lukas Wagner <l.wagner@proxmox.com>
This commits also makes user authentication async, so that e.g. a not
responding LDAP server cannot block other logins.
Signed-off-by: Lukas Wagner <l.wagner@proxmox.com>
Note: bind-passwords set via the API are not stored in `domains.cfg`,
but in a separate `ldap_passwords.json` file located in
`/etc/proxmox-backup/`.
Similar to the already existing `shadow.json`, the file is
stored with 0600 permissions and is owned by root.
Signed-off-by: Lukas Wagner <l.wagner@proxmox.com>
The function was moved to proxmox-rest-server to make it
usable in the proxmox-backup-manager cli binary.
Signed-off-by: Lukas Wagner <l.wagner@proxmox.com>
Running configured jobs was already possible using the Web UI, but not
using the CLI. To fix that, this commit adds the following commands to
`proxmox-backup-manager`:
* prune-job run <id>
* sync-job run <id>
* verify-job run <id>
Signed-off-by: Friedrich Weber <f.weber@proxmox.com>
instead of
Error: unable to open snapshot directory "/full/path/to/snapshot" for locking - ENOENT: No such file or directory
this will now print
Error: Snapshot vm/800/2023-01-16T12:28:11Z does not exist.
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
In former commit, the enum members were renamed to be CamelCase, in
accordance with the usual Rust style guide. However, this broke the
GUI in some places due to failing JSON property deserialization.
To fix this, some serde(rename = "kebab-case") directives were added.
Some properties were also serde-renamed to snake_case, otherwise
it would have been necessary to also modify proxmox-widget-toolkit
as well as PVE source code. This can follow in a later commit if so
desired.
Fixes: a2055c38 fix non-camel-case enums
Signed-off-by: Lukas Wagner <l.wagner@proxmox.com>
`proxmox-backup-client task log ..` and `proxmox-backup-manager task log ..`
are used to view the logs of tasks that have been started by another client, so
interrupting the task progress view should not forward the interrupt to the
running task. other call sites of the same helper(s) that spawn a task and then
print its progress should keep the old behaviour of interrupting the spawned
task on C^c.
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
Was a bit odd that the very similar (same return type) methods used
a for+if/else+push and a iter+filter+collect approach.
Switch both to the latter and use a match instead of if/else for
shorter code
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
while refactoring the the empty media set checks, we accidentally
reversed one check from !is_empty to unassigned, which now never
included the right media sets.
reversing the condition fixes that.
fixes
52517f7b: ("tape: hide internal use of all zero uuid for unassigned tapes")
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
