6 messages in com.mysql.lists.clusterRe: Storage node groups| From | Sent On | Attachments |
|---|---|---|
| Aslan Carlos | 20 Mar 2007 14:06 | |
| Alex Davies | 20 Mar 2007 19:45 | |
| Christiaan den Besten | 21 Mar 2007 04:03 | |
| Aslan Carlos | 21 Mar 2007 11:37 | |
| Alex Davies | 21 Mar 2007 17:46 | |
| Aslan Carlos | 21 Mar 2007 20:02 |
| Subject: | Re: Storage node groups |
|---|---|
| From: | Alex Davies (al...@davz.net) |
| Date: | 03/21/2007 05:46:25 PM |
| List: | com.mysql.lists.cluster |
Hi,
I dont think I quite understand what you are trying to do.
Fundermentally, a cluster of 22 nodes is going to have to partition (split up) your data over those 22 nodes. Therefore, not each node will contain all the data. By extension, you will have "storage groups" - and if every node in a group fails, you will get a total failure.
However, if you set the number of replicas to be say 4 that means 4 machines must fail (and it must be 4 in the same group - statistically unlikely) to take the cluster down. Is this acceptable to you?
Alex
On 21/03/07, Aslan Carlos <asl...@digirati.com.br> wrote:
thanks Alex,
I try this , but cluster scalability will be equal 0, because i have =~ 22 servers for NDB to build my cluster, so the number of replicas will be 8 ( if I'm not wrong, this values is the top of Nofreplicas).
So 8 Replicas / 22 Machines =~ 2.75 groups, so I can put more 2 machines in the cluster, but the cluster will make 3 storage group, and I can't put one machine in the future , just 8 server together in the cluster.
The scalability is crucial for my cluster. :( and this problem makes a sound not nice for my project. :(
So anyone has other hit?
Thanks in advanced.
-- Aslan Carlos de M. Ramos Network Administrator K8 Networks, Digirati Networks Tel: +55(21)2233-5950 Ramal:29 Skype: aslancarlos
On Qua, 2007-03-21 at 02:46 +0000, Alex Davies wrote:
Hi,
I think I understand what you are asking.
To ensure that all nodes have all data, you simply have to ensure that the number of replicas is equal to the number of nodes - i.e. 2 replicas in a 2 node cluster. In this scenario, both nodes will hold 100% of the data and either can fail without a problem.
This however will remove some of the performance advantage of a cluster, and will realistically prevent you from storing massive amounts of data in a cost efficient way.
Alex
On 20/03/07, Aslan Carlos <asl...@digirati.com.br> wrote: Hi Community,
I have a question about the node storage group..
I did some tests which I removed all nodes in the same storage group, but the cluster stop when I did this. So my nodes group have 2 node storage ( replica 2 ) .
There a way to do the cluster don't stop when it lost one storage group?
For example: I'm using shared-nothing nodes, well if 2 nodes has some problem, maybe hardware fail, and they is in the same storage group, my cluster go down. so the HA fail in this point.
This don't make sense for me and I don't found any answer about this in nowhere.
How could I make one big group , where all nodes has 100% of data fragments?
Sorry, English isn't my natural language, if anybody don't understand, I'll try to be more clear.
Thanks in avanced. -- Aslan Carlos de M. Ramos Network Administrator K8 Networks, Digirati Networks Tel: +55(21)2233-5950 Ramal:29 Skype: aslancarlos
-- Alex Davies // http://www.davz.net
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the sender immediately by e-mail and delete this e-mail permanently.
-- Alex Davies // http://www.davz.net
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the sender immediately by e-mail and delete this e-mail permanently.