1

Has anyone experience of doing full VM snapshot or rsync back-ups with Plone? We have a server running several separate Zope application server worth of around 10-20 GB ZODB data and blobs.

  • Would it be possible to duplicate the server by rsynching the whole server filesystem to a replicate machine once in a day?

This would provide "barebone high availability" where in the case of server breakage one could just switch the production IP to the back-up server.

The issues here are

  • Can rsync delta algorithm cleverly copy append-only Data.fs files or does it copy the whole Data.fs every time? Data.fs and blobs are the only data changing on the server running Plone sites.

  • Is rsync fast enough for calculating deltas for the whole server file system once in a day

  • Does rsync cause too much stress to the system when it is running

  • Do one need to run something more advanced over rsync to speed up the process (e.g. Duplicity)

  • Does Ubuntu system partition has something which must be excluded from rsync

As far as I know live ZODB Data.fs copies are ok due to their append-only nature.

The benefits of such replication method would

  • Very simple to set-up: No advanced replication knowledge beyond normal Linux sysadmin knowledge needed

The data in our case is not that critical. We are fine if we can restore 1-2 old back-ups as long as the replication takes over within one day.

The server is running Ubuntu 12.04.

4

1 回答 1

4

是的，rsync 将能够以增量方式复制 Data.fs 数据；毕竟，这是一种仅附加的方法。

当被 rsync 复制时，你确实冒着最后一个正在进行的事务不完整的风险，但是 Zope 足够聪明，可以忽略这样的不完整事务。

rsync 将不得不做很多工作来找出在 Data.fs 中从哪里开始复制，因为它不知道格式只是附加的。所以在每次运行时，它都会从头开始分析整个文件，这会消耗一些 CPU 和磁盘访问资源。

也许使用repozo和 rsync 的组合会是一个更好的解决方案？运行 repozo 以创建快照（创建快照效率更高，因为 repozo确实知道 Data.fs 的仅附加性质），重新同步这些快照，并自动从这些备份中“恢复”您的冷故障转移服务器。在collective.recipe.backup你的构建中包含 repozo 是一件轻而易举的事。

于 2012-08-02T12:07:04.287 回答