[MDEV-6228] Mariadb galera cluster slave lag Created: 2014-05-11  Updated: 2015-06-12  Resolved: 2014-05-11

Status: Closed
Project: MariaDB Server
Component/s: None
Affects Version/s: 5.5.37-galera
Fix Version/s: None

Type: Bug Priority: Major
Reporter: QFY Assignee: Unassigned
Resolution: Not a Bug Votes: 0
Labels: None
Environment:

CentOs 6.2
virtual PC ,use 'bridge model' to connect network.


 Description   

When Mariadb galera cluster comes across a big transaction ,like insert into a table 1000,000 rows ,or update a table 1000,000 rows ,Doesn't the slave
lag?
when I commit at the first node with a big transaction, and It show complete. But the other nodes don't have the same date with the first one and after some seconds they do.
The galera cluster just copy the transaction to the other nodes,and say it complete.So mariadb glare cluster miss a transaction it have lagged,but when the transaction is small ,the time it used is small ,and we doesn't fell it.

 Comments   
Comment by Elena Stepanova [ 2014-05-11 ]

Please read documentation on Galera replication, in particular how certification-based replication works (e.g. here: http://galeracluster.com/documentation-webpages/certificationbasedreplication.html ), but reading other sections can be useful too.

If after doing so you still think you observe a bug, please comment to re-open the issue, attach logs and any other materials that can help to confirm and analyze the problem.

Comment by QFY [ 2014-05-12 ]

I mean that Mariadb galera cluster said there is no slave lag.
So at any given time, any node have the same data .And that is said 'no slave lag'.
Actually there is having a slave lag , and the period is the transaction that runs in slave.

I just mean this.Maybe it's not a bug (actually it is not a bug),but the document in official must be preciseness.

The official document :https://mariadb.com/kb/en/what-is-mariadb-galera-cluster/

Comment by Frens Jan Rumph [ 2015-06-12 ]

(sorry for commenting on an old ticket, but I reckoned it better than opening a new one)

About the last comment by @QFY ... is this correct? If an (commit for a) insert / update acknowledged, can a reader on another node read stale data?

Generated at Thu Feb 08 07:10:18 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.