If a fuse_loop instance dies suddenly (e.g. SIGKILL), the FUSE mount and
loop device assignment are left behind. We can determine this scenario
on specific unmap, when the PID file is either missing or contains a PID
of a non-running process, but the backing file and potentially loop
device are still there.
If that's the case, do an "emergency cleanup", by unassigning the
loopdev, calling 'fusermount -u' and then cleaning any leftover files
manually.
With this in place, pretty much any situation is now recoverable via
only the 'proxmox-backup-client' binary, by either calling 'unmap' with
or without parameters.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
On unmap, only report success if the instance we are killing actually
terminates. This is especially important so that cleanup routines can be
assured that /run files are actually cleaned up after calling
cleanup_unused_run_files.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
A 'map' call will only clean up what it needs, that is only leftover
files or dangling instances of it's own name.
For a full cleanup the user can call 'unmap' without any arguments.
The 'cleanup on error' behaviour of map_loop is removed. It is no longer
needed (since the next call will clean up anyway), and in fact fixes a
bug where trying to map an image twice would result in an error, but
also cleanup the .pid file of the running instance, causing 'unmap' to
fail afterwards.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
So user doesn't need to remember which loop devices he has mapped to
what.
systemd unit encoding is used to transform a unique identifier for the
mapped image into a suitable name. The files created in /run/pbs-loopdev
will be named accordingly.
The encoding all happens outside fuse_loop.rs, so the fuse_loop module
does not need to care about encodings - it can always assume a name is a
valid filename.
'unmap' without parameter displays all current mappings. It's
autocompletion handler will list the names of all currently mapped
images for easy selection. Unmap by /dev/loopX or loopdev number is
maintained, as those can be distinguished from mapping names.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Allows mapping fixed-index .img files (usually from VM backups) to be
mapped to a local loopback device.
The architecture uses a FUSE-backed temp file mapped to a loopdev:
/dev/loopX -> FUSE /run/pbs-loopdev/xxx -> backup client -> PBS
Since unmapping requires some cleanup (unmap the loopdev, stop FUSE,
remove the temp files) a special 'unmap' command is added, which uses a
PID file to send SIGINT to the backup-client instance started with
'map', which will handle the cleanup itself.
The polling with select! in mount.rs needs to be split in two, since we
have a chicken and egg problem between running FUSE and setting up the
loop device - so we need to do them concurrently, until the loopdev is
assigned, at which point we can report success and daemonize, and then
continue polling the FUSE loop future.
A loopdev module is added to tools containing all required functions for
mapping a loop device to the FUSE file, with the ioctls moved into an
inline module to avoid exposing them directly.
The client code is placed in the 'mount' module, which, while
admittedly a loose fit, allows reuse of the daemonizing code.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Causes a panic if last_update is smaller than RRD_DATA_ENTRIES*reso,
which (I believe) can happen when inserting the first value for a DB.
Clamp the value to 0 in that case.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Fix a potential bug where errors that happen after the SendHandle has
been dropped while doing the thread join might have been ignored.
Requires internal check_abort to be moved out of 'impl SendHandle' since
we only have the Mutex left, not the SendHandle.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
A client can omit uploading chunks in the "known_chunks" list, those
then also won't be written on the server side. Check all those chunks
mentioned in the index but not uploaded for existance and report an
error if they don't exist instead of marking a potentially broken backup
as "successful".
This is only important if the base snapshot references corrupted chunks,
but has not been negatively verified. Also, it is important to only
verify this at the end, *after* all index writers are closed, since only
then can it be guaranteed that no GC will sweep referenced chunks away.
If a chunk is found missing, also mark the previous backup with a
verification failure, since we know the missing chunk has to referenced
in it (only way it could have been inserted to known_chunks with
checked=false). This has the benefit of automatically doing a
full-upload backup if the user attempts to retry after seeing the new
error, instead of requiring a manual verify or forget.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Do not allow clients to reuse chunks from the previous backup if it has
a failed validation result. This would result in a new "successful"
backup that potentially references broken chunks.
If the previous backup has not been verified, assume it is fine and
continue on.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
We need to update the atime of chunk files if they already exist,
otherwise a concurrently running GC could sweep them away.
This is protected with ChunkStore.mutex, so the fstat/unlink does not
race with touching.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
The iterator of get_chunk_iterator is extended with a third parameter
indicating whether the current file is a chunk (false) or a .bad file
(true).
Count their sizes to the total of removed bytes, since it also frees
disk space.
.bad files are only deleted if the corresponding chunk exists, i.e. has
been rewritten. Otherwise we might delete data only marked bad because
of transient errors.
While at it, also clean up and use nix::unistd::unlinkat instead of
unsafe libc calls.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
This ensures that following backups will always upload the chunk,
thereby replacing it with a correct version again.
Format for renaming is <digest>.<counter>.bad where <counter> is used if
a chunk is found to be bad again before a GC cleans it up.
Care has been taken to deliberately only rename a chunk in conditions
where it is guaranteed to be an error in the chunk itself. Otherwise a
broken index file could lead to an unwanted mass-rename of chunks.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Anyone with a PAM account and Sys.Console access could have started a
termproxy session, adapt the regex.
Always test for broken entries and run the sed expression to make sure
eventually all occurences of the broken syntax are fixed.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
To prevent forgetting the base snapshot of a running backup, and catch
the case when it still happens (e.g. via manual rm) to at least error
out instead of storing a potentially invalid backup.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
This reverts commit d53fbe2474.
The HashSet and "register" function are unnecessary, as we already know
which backup is the one we need to check: the last one, stored as
'last_backup'.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
An flock on the snapshot dir itself is used in addition to the group dir
lock. The lock is used to avoid races with forget and prune, while
having more granularity than the group lock (i.e. the group lock is
necessary to prevent more than one backup per group, but the snapshot
lock still allows backups unrelated to the currently running to be
forgotten/pruned).
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Attempt to lock the backup directory to be deleted, if it works keep the
lock until the deletion is complete. This way we ensure that no other
locking operation (e.g. using a snapshot as base for another backup) can
happen concurrently.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
If the datastore holds broken backups for some reason, do not attempt to
base following snapshots on those. This would lead to an error on
/previous, leaving the client no choice but to upload all chunks, even
though there might be potential for incremental savings.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Commit 9fa55e09 "finish_backup: test/verify manifest at server side"
moved the finished-marking above some checks, which means if those fail
the backup would still be marked as successful on the server.
Revert that part and comment the line for the future.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
This should never trigger if everything else works correctly, but it is
still a very cheap check to avoid wrongly marking a backup as "OK" when
in fact some chunks might be missing.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Multiple backups within one backup group don't really make sense, but
break all sorts of guarantees (e.g. a second backup started after a
first would use a "known-chunks" list from the previous unfinished one,
which would be empty - but using the list from the last finished one is
not a fix either, as that one could be deleted or pruned once the first
simultaneous backup is finished).
Fix it by only allowing one backup per backup group at one time. This is
done via a flock on the backup group directory, thus remaining intact
even after a reload.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
To prevent a race with a background GC operation, do not allow deletion
of backups who's index might currently be referenced as the "known chunk
list" for successive backups. Otherwise the GC could delete chunks it
thinks are no longer referenced, while at the same time telling the
client that it doesn't need to upload said chunks because they already
exist.
Additionally, prevent deletion of whole backup groups, if there are
snapshots contained that appear to be currently in-progress. This is
currently unlikely to trigger, as that function is only used for sync
jobs, but it's a useful safeguard either way.
Deleting a single snapshot has a 'force' parameter, which is necessary
to allow deleting incomplete snapshots on an aborted backup. Pruning
also sets force=true to avoid the check, since it calculates which
snapshots to keep on its own.
To avoid code duplication, the is_finished method is factored out.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
...when an entry is selected, that doesn't exist after the reload.
E.g. when one deletes selects a file within a snapshot and then clicks
the delete icon for said snapshot, focusRow would then fail and the
loading mask stay on until a reload.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Depends on patched apt-pkg-native-rs. Changelog-URL detection is
inspired by PVE perl code for now, though marked with fixme to use 'apt
changelog' later on, if/when our repos have APT-compatible changelogs
set up.
list_installed_apt_packages iterates all packages and creates an
APTUpdateInfo with detailed information for every package matched by the
given filter Fn.
Sadly, libapt-pkg has some questionable design choices regarding their
use of 'iterators', which means quite a bit of nesting...
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Requires updating the AsyncRead implementation to cope with byte-wise
seeks to intra-chunk positions.
Uses chunk_from_offset to get locations within chunks, but tries to
avoid it for sequential read to not reduce performance from before.
AsyncSeek needs to use the temporary seek_to_pos to avoid changing the
position in case an invalid seek is given and it needs to error in
poll_complete.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
This aligns it with PVE and allows the widget toolkit's update window
"refresh" to work without modifications once POST /apt/update is
implemented.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
To support incremental backups (where not all chunks are sent to the
server), a new parameter "reuse-csum" is introduced on the
"create_fixed_index" API call. When set and equal to last backups'
checksum, the backup writer clones the data from the last index of this
archive file, and only updates chunks it actually receives.
In incremental mode some checks usually done on closing an index cannot
be made, since they would be inaccurate.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
if *only* data chunks are registered (high chance during incremental
backup), then chunk_count might be one lower then upload_stat.count
because of the zero chunk being unconditionally uploaded but not used.
Thus when subtracting the two, an overflow would occur.
In general, don't let the client make the server panic, instead just set
duplicates to 0.
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
