简体繁体 English

在其他percona xtradb群集节点上重用EBS快照

[英]Reusing EBS snapshots on a different percona xtradb cluster node

原文 2015-04-28 12:39:49 5 1 mysql/ amazon-web-services/ amazon-ec2/ percona

I'm evaluating a Percona xtradb 5.6 cluster of 3 nodes in AWS environment. 我正在评估AWS环境中3个节点的Percona xtradb 5.6集群。 I'm using ec2-consistent-snapshot with --mysql to make an EBS snapshot of the data. 我将ec2-consistent-snapshot与--mysql配合--mysql来制作数据的EBS快照。 However when a snapshot was made on node 1 , and then node 2 is relaunched using that snapshot, the cluster would break. 但是，当在node 1上创建快照，然后使用该快照重新启动node 2 ，群集将损坏。

Through trial-and-error I've found that this is caused by reusing auto.cnf and gvwstate.dat files in mysql datadir, which would contain ids of node 1 , and the issues were (apparently) caused by another node trying to join with id of another node already in cluster. 通过反复试验，我发现这是由于在mysql datadir中重用了auto.cnf和gvwstate.dat文件引起的，该文件将包含node 1 ID，而这些问题（显然）是由另一个试图加入的节点引起的集群中已经有另一个节点的ID。 Removing the said files appears to have fixed the issue and now nodes go up and down as expected. 删除上述文件似乎已解决了该问题，现在节点可以按预期的方式上下移动。

My question is: did I do the right thing? 我的问题是：我做对了吗？ Do I need to remove auto.cnf and gvwstate.dat before using another server's datadir? 使用其他服务器的数据目录之前，是否需要删除auto.cnf和gvwstate.dat？ Do I need to do anything else? 我还需要做其他事情吗？ What's the standard practice for this sort of thing? 这种事情的标准做法是什么？

1 个解决方案

What you did was correct. 你所做的是正确的。 However, be sure to check your gcache size to avoid SST. 但是，请务必检查您的gcache大小以避免SST。 It is quite possible that you could take the ebs-snap now, go to lunch, come back and create node3 using that snap, start mysql up and an SST happens anyway. 很可能您现在可以立即进行ebs快照，吃午餐，返回并使用该快照创建node3，启动mysql并无论如何都会发生SST。

I would check your logs on the new node to ensure that an SST did NOT happen. 我将检查您在新节点上的日志，以确保未发生SST。