You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@ignite.apache.org by 38797715 <38...@qq.com> on 2021/05/10 08:28:27 UTC
Data synchronization after node restart
Hi team,
If persistence is enabled and the number of backups is 1.
At this time, if a node fails, and the data writing is still normal.
If the previously failed node restarts, it joins the cluster again.
So is this restored node synchronizing all data from other nodes or only
the differential data during failure?
Re: Data synchronization after node restart
Posted by akorensh <al...@gmail.com>.
The case you are referring to is a baseline node being offline for a period
of time.
Like I mentioned before, once this node comes back online it will undergo
historical rebalancing to
catch on data it missed.
like you said, idle_verify is a tool used to determine if there are
inconsistencies between backup and primary, but is not related to the use
case presented. See:
https://ignite.apache.org/docs/latest/tools/control-script#cluster-should-be-idle-during-idle_verify-check
Each node writes to the WAL on its own in order to prevent loss of data.
https://ignite.apache.org/docs/latest/persistence/native-persistence.html#checkpointing
During the aforementioned historical rebalancing process, the WAL is being
used to determine the delta between the data the previously offline node
has and what it needs to have.
https://www.gridgain.com/docs/latest/developers-guide/historical-rebalancing
from the doc:
Historical rebalancing relies on the history of updates stored in the
write-ahead-log files (WALs) of cluster nodes. The WALs' history must keep
updates of all the partitions which deltas need to be rebalanced to a
restarted node.
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
Re: Data synchronization after node restart
Posted by 38797715 <38...@qq.com>.
Hello,
We know that the control script has an idle_verify command can be used
to verify the potential inconsistency between the primary and backup.
Well, since there are WAL and WAL archives, they can be used for
historical data balancing. So why can't WAL ensure the consistency
between the primary and the backup? Does the backup update also need to
write the WAL file?
在 2021/5/10 下午9:57, akorensh 写道:
> Hi,
> You are referring to a persistent node failing while being a part of a
> baseline topology.
> When that same node comes back, it will load only the delta(differential
> per your definition) from
> the time that it was down. This is called historical rebalancing.
> Read more here:
> https://www.gridgain.com/docs/latest/developers-guide/historical-rebalancing
> More on baseline topology:
> https://ignite.apache.org/docs/latest/clustering/baseline-topology
> Thanks, Alex
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/
Re: Data synchronization after node restart
Posted by akorensh <al...@gmail.com>.
Hi,
You are referring to a persistent node failing while being a part of a
baseline topology.
When that same node comes back, it will load only the delta(differential
per your definition) from
the time that it was down. This is called historical rebalancing.
Read more here:
https://www.gridgain.com/docs/latest/developers-guide/historical-rebalancing
More on baseline topology:
https://ignite.apache.org/docs/latest/clustering/baseline-topology
Thanks, Alex
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/