Borg wrappers for server pull mode.
As distributions repositories may not always be updated, it is designed to work with the standalone version, installed in /opt/bbbs
.
borg serve
process is not reused, 1 backup = 1 client = 1 wrapper call = 1 serveknown_hosts
) keyUTF-8
) localeOne option is to git clone
this repo in /opt/bbbs/
and run the installer.\
Obviously, you have to create do-backup
from do-backup.example
according to your needs and mark it as executable (chmod +x /opt/bbbs/do-backup
).\
borg's dot path are symlinked to more accessible paths (.cache/borg -> ~/cache
, .config/borg -> ~/
(to expose keys
and security
directly), .ssh -> ssh
).
client> apt-get update && apt-get install git sudo socat curl jq
client> git clone https://github.com/u1735067/bbbs.git /opt/bbbs
client> /opt/bbbs/installer.sh install-client
server> yum install git sudo socat curl jq
server> git clone https://github.com/u1735067/bbbs.git /opt/bbbs
server> /opt/bbbs/installer.sh install-server
server> /opt/bbbs/ssh-gen-copy-key --generate --key-name server-name_hypervisor --copy --user borg -- root@server-name -p 22
server> gw_key=~borg/ssh/key_server-name_hypervisor; /opt/bbbs/ssh-gen-copy-key --generate --key-name server-name_vm-1 --copy --user borg -- root@172.16.0.1 -p 22 -o ProxyCommand="ssh -i$gw_key -W %h:%p server-name.fqdn -p 22"
server> /opt/bbbs/borg init -e authenticated-blake2 /srv/borg/server-name/hypervisor
server> /opt/bbbs/borg init -e authenticated-blake2 /srv/borg/server-name/vm-1
server> sudo -u borg /opt/bbbs/do-backup
server> /opt/bbbs/borg info /srv/borg/server-name/hypervisor
bbbs-client
will call ~borg/backup-pre
and ~borg/backup-pre
on the client before and after running borg create
, you can use them to dump a sql database and remove it for example.
To manage ssh connexion parameters, you can also use ssh_config
instead, for example:
server> cat ~borg/ssh/config
Host *
User borg
Host server-name
Port 22
IdentityFile ~borg/ssh/key_server-name_hypervisor
Host 172.16.*.*
ProxyCommand ssh -F ~borg/ssh/config -W %h:%p server-name
╔══════╦════════════════════════════════════════════════════╦══════════════════════════════════════════════════════════════╗
║ ║ Server ║ Client ║
╠══════╬════════════════════════════════════════════════════╬══════════════════════════════════════════════════════════════╣
║ root ║ sudo ║ bbbs-client ─┬─► ~borg/backup-pre ║
║ ║ │ ║ ▲ ├─► borg create ─► wrapper ─► socat ║
║ ║ │ ║ │ └─► ~borg/backup-post │ ║
║ ║ │ ║ │ │ ║
╠══════╬════┼═══════════════════════════════════════════════╬══════════════┼══════════════════════════════════════════╪════╣
║ borg ║ └─► do-backup ║ │ │ ║
║ ║ └─► wrapper ║ │ │ ║
║ ║ ├──────────────────────► ssh ====╬===► ssh ──► sudo │ ║
║ ║ └─► socat ──► borg serve ┊ ║ ┊ │ ║
║ ║ ▲ ┊ ║ ┊ │ ║
║ ║ └──────────────────╘======╬======╛◄─────────────────────────────────────────────────┘ ║
╚══════╩════════════════════════════════════════════════════╩══════════════════════════════════════════════════════════════╝
===== = ssh channel
*/!\ Don't forget to set executable bit (`chmod +x ~borg/backup-p`) on them!**
Cache credentials in ~borg/.mylogin.cnf
(see mysqldump
, mysql_config_editor
):
su borg -c "mysql_config_editor set --user=root --password"
or for MariaDB
in ~borg/.my.cnf
(see mysqldump
):
[mysqldump]
user=mysqluser
password=secret
~borg/backup-pre
:
#!/usr/bin/env bash
echo "- Dumping MySQL ..."
mkdir --parents /var/backups/mysql
# GZip directly in case the database is huge -- could be handled by borg using stdin maybe
mysqldump --all-databases --single-transaction \
| gzip --fast --rsyncable \
> /var/backups/mysql/borg-dump_$(date --utc "+%Y-%m-%d_%H.%M.%SZ").sql.gz
echo "- Dumping MySQL ... Done"
~borg/backup-post
:
#!/usr/bin/env bash
echo "- Removing MySQL Dump ..."
rm -f /var/backups/mysql/borg-*
echo "- Removing MySQL Dump ... Done"
See Backup indexed data, Index backup strategy.
~borg/backup-pre
:
#!/usr/bin/env bash
echo "- Dumping Splunk ..."
mkdir --parents /var/backups/splunk
/opt/splunk/bin/splunk search "index=*" -output json -maxout 0 \
| gzip --fast --rsyncable \
> /var/backups/splunk/borg-dump_$(date --utc "+%Y-%m-%d_%H.%M.%SZ").json.gz
echo "- Dumping Splunk ... Done"
echo "- Stopping Splunk ..."
service splunk stop
#systemctl stop splunk
echo "- Stopping Splunk ... Done"
~borg/backup-post
:
#!/usr/bin/env bash
echo "- Removing Splunk Dump ..."
rm -f /var/backups/splunk/borg-*
echo "- Removing Splunk Dump ... Done"
echo "- Starting Splunk ..."
service splunk start # || mail failed to restart service, check ASAP
#systemctl start splunk
echo "- Starting Splunk ... Done"
While Borg Backup Server
(BBS) has been announced, it is not available yet and I needed a solution (preferably before my server crashes). Also, I don't want the clients to have to know how to reach the server.
This is ugly but probably the simplest for now (when using pull mode).
A correct solution could be to use python
as wrapper, with libssh
to setup the channel, and parsing nice configuration files.
The wrappers borg
and bbbs-client
will try to set a correct locale (meaning: with UTF-8
charset).
Because Python relies on it to read filenames correctly. Most Unix filesystems seems to only care about bytes
, but the convention is to use UTF-8 and most locales without the charset specifier (.utf8
) doesn't not use UTF-8, so Python use that charset and fails to decode filenames properly.
Also, distributions like OMV can decide to set LANG=C
before executing crons, which will make Python use the ascii
/ANSI
encoding.
However, except for extract
/ fuse mouse
according to the issues, borg seems to handle that correctly (even though I'm not sure why, as args.path
are str
, so scandir will return str
which means they are decoded?).
More informations (about this mess):
To do so, it'll parse locales by priority (LC_ALL
> LC_CTYPE
> LANG
), testing if it's indicating UTF-8
charset, or if it's not trying to find the UTF-8
alternative of this locale, or ultimately unsetting it (to let the chance to lower priority locales).\
If there's no locale left, it'll then try to see if there's any UTF-8
locale available, setting LC_CTYPE
, starting by C
, English (en_GB
, en_US
, en_*
) and ultimatly the first found.\
In the end, if there's really no UTF-8
locale available, none of LC_ALL
, LC_CTYPE
, LANG
should be set, thus triggering the python 3 default UTF-8
, but I wasn't able to reproduce this behavior:
# export -n LC_ALL LC_CTYPE LANG; python3 -c 'import sys; print(sys.version); print(sys.getfilesystemencoding())'
3.4.2 (default, Oct 8 2014, 10:45:20)
[GCC 4.9.1]
ascii