[pbs-devel] [PATCH vma-to-pbs v6 0/4] add support for bulk import of a dump directory

Filip Schauer f.schauer at proxmox.com
Wed Nov 13 16:57:58 CET 2024


When a path to a directory is provided in the vma_file argument, try to
upload all VMA backups in the directory. This also handles compressed
VMA files, notes and logs. If a vmid is specified with --vmid, only the
backups of that particular vmid are uploaded.

Also improve the readability of the log messages to keep track of all
imported backups.

Changed since v5:
* Extend the help text by seperate usages for single VMA import and bulk
  import
* Move variables into format strings where possible

Changed since v4:
* Switch grouped_vmas from Vec<Vec<>> to HashMap<String, Vec<>>
* Remove dependency on itertools
* bail when no backups were found
* Default to yes on the bulk import confirmation prompt
* bail on invalid input to the bulk import confirmation prompt

Changed since v3:
* Mention in the description of the --vmid argument, that it is required
  if a single VMA file is provided
* Construct grouped_vmas in place
* Add debug logs when gathering files for bulk import
* Log a summary of the files gathered for bulk import
* Remove the "confusing VMA file path" error message in the second
  commit
* Switch chunk_stats from Arc<Mutex<[u64; 256]>> to
  Arc<[AtomicU64; 256]> and use fetch_add to atomically increment and
  fetch the chunk stat
* Ask for confirmation before bulk import
* Add --yes option to skip the confirmation prompt

Changed since v2:
* Make skipping a VMID on error optional with the --skip-failed option
* Switch log output from stderr to stdout
* Bump itertools to 0.13

Changed since v1:
* Do not recurse through dump directory
* Compile regex once before iterating over the files in the dump
  directory
* Use extract on regex capture groups
* Do not use deprecated method `chrono::NaiveDateTime::timestamp`
* Use proxmox_sys::fs::file_read_optional_string
* Group VMA files by VMID and continue with next VMID on error
* Move the BackupVmaToPbsArgs split into its own commit
* Remove hard coded occurences of 255
* Use level-based logging instead of println

Filip Schauer (4):
  add support for bulk import of a dump directory
  add option to skip vmids whose backups failed to upload
  use level-based logging instead of println
  log device upload progress as a percentage

 Cargo.toml     |   4 +
 src/main.rs    | 198 +++++++++++++++++++++++++++++++++++++++++++++----
 src/vma2pbs.rs | 108 +++++++++++++++++++++------
 3 files changed, 271 insertions(+), 39 deletions(-)

-- 
2.39.5





More information about the pbs-devel mailing list