I think you can avoid this , eg : A, B, C is seed node and D, E is non-seed. A crash and damage.
1) re-configure the node A to a non-seed; 2) starting the non-seed nood A will re-bootstrap; 3) Though the original seed nodes will decrease 1, any other node, eg D, can re-configure other nodes, eg B, C, E, as seed nodes. 4) Then the seed nodes in the ring will be : Node->Seeds List A --> B, C, D; B --> A, B, C; C --> A, B, C; D --> B, C, E; E --> A, B, C; Am I right?? ;-) ------------------ XL.Pan 2009-12-23 ------------------------------------------------------------- 发件人：mail.list.steel.mental 发送日期：2009-12-23 17:43:29 收件人：[email protected] 抄送： 主题：RE: Why seed can't startup in Bootstrap mode ? So, once a seed crash and damage it's data, we cannot wipe the node and re-bootstrap, because node cannot bootstrap if configured as seed. Also we can not re-configure node to a non-seed, because that will reduce total number of seeds, so eventually all old data will lost(one by one), or there is no seed(all seed was re-configure to non-seed). Am I right?? ---------END---------- -----Original Message----- From: Jonathan Ellis [mailto:[email protected]] Sent: Wednesday, December 23, 2009 11:46 AM To: [email protected] Subject: Re: Why seed can't startup in Bootstrap mode ? On Tue, Dec 22, 2009 at 9:39 PM, zhangyf2007 <[email protected]> wrote: > Consider the following scene: > 1) the ReplicationFactor is 3 > 2) the cluster has 3 nodes and all the nodes are seed nodes > 3) one node is dead while the data it managed is also damaged unfortunately > 4) it starts up but it can't request any data from other node you can either repair the existing data or wipe the node, remove it from the cluster, and rebootstrap. this has nothing to do with being seeds; bootstrapping is for adding new nodes to the cluster, doing it to a node that already has data on it is wrong. see http://wiki.apache.org/cassandra/Operations -Jonathan
