ZFS snapshot replication tool
autosync
will scan ZFS datasets on source server and automatically mirror
all enabled by include/exclude rules datasets to destination dataset on local server.
cd /opt
git clone https://github.com/makhomed/autosync.git autosync
Also you need to install python3:
# yum install python3
cd /opt/autosync
git pull
vim /opt/autosync/source-server.conf
- write to config something like this:
source source-server.example.com
exclude tank
destination tank/mirror/source-server.example.com
Configuration file allow comments, from symbol #
to end of line.
Configuration file has only six directives:
source
, exclude
, include
, destination
, save
and delay
.
Syntax of source
directive: source <source-server>[:port]
.
<source-server>
is hostname of source server or it ip address.
Port is optional.
Syntax of include
and exclude
directives are the same:
exclude <pattern>
or include <pattern>
.
By default all datasets are included. But you can exclude some datasets
by name or by pattern. Pattern is rsync-like, ?
means any one symbol,
*
means any symbols except /
symbol, **
means any symbols.
First match win, and if it was directive exclude
- dataset will be excluded,
if it was directive include
- dataset will be included.
exclude
and include
define datasets for replication from source server.
destination
directive define destination dataset name on the current server.
Syntax of save
directive: save <interval> <count>
. For example:
save hourly 24
save daily 30
save weekly 8
save
directive can be global - for all datasets by default, or local, for specific dataset.
For example:
source example.com
exclude tank
destination tank/mirror/example.com
save hourly 24
save daily 30
save weekly 8
[tank/mirror/example.com/kvm-stage-elastic]
save hourly 24
save daily 15
save weekly 8
[tank/mirror/example.com/kvm-stage-mysqld]
save hourly 24
save daily 15
save weekly 8
By default, if no directive save
exists for specific interval, 1:1 replica will be created,
and all snapshots, not existent on source server, will be deleted on local server for destination datasets.
delay
defines delay in seconds between two sequential run of sync. Default value is 600 seconds.
For work you need to generate private ssh key on destination server
with comamnd ssh-keygen -t rsa
and copy public key from /root/.ssh/id_rsa.pub
to /root/.ssh/authorized_keys
on source servers. Also you need to check connection
with command ssh source-server.example.com
and answer yes
on ssh question:
# ssh source-server.example.com
The authenticity of host 'source-server.example.com' can't be established.
ECDSA key fingerprint is SHA256:/cYI0bJzEX+CF3DhGEUQ+ZeGFmMzEJYAt3C15450zKs.
ECDSA key fingerprint is MD5:44:20:bd:f5:aa:a7:52:ac:c5:19:e5:e0:28:2b:90:49.
Are you sure you want to continue connecting (yes/no)? yes
vim /etc/systemd/system/[email protected]
- write to unit file something like this:
[Unit]
Description=autosync %I
After=network-online.target
[Service]
ExecStart=/opt/autosync/autosync -c /opt/autosync/%i.conf
Restart=always
RestartSec=60s
StartLimitInterval=0
[Install]
WantedBy=multi-user.target
Note: in new versions of systemd StartLimitInterval renamed to StartLimitIntervalSec and moved from [Service] to [Unit] section. See details at https://selivan.github.io/2017/12/30/systemd-serice-always-restart.html
After this you need to start service:
systemctl daemon-reload
systemctl enable autosync@source-server
systemctl start autosync@source-server
systemctl status autosync@source-server
If all ok you will see what service is enabled and running.
Details about replication process you can seee in the log files in the log directory.