SUN Custer Theory

UNIX Solaris
1

Hi,

Newbie to the cluster, Would like to some basic idea on the cluster framework.:confused:

Any pointer to a link which is useful.:stuck_out_tongue:

How to explain on cluster's disk? :frowning:

Scenario, node1, node2, Storage1 and Storage2.

  1. If storage1 attached to node1, and node1 fault; node1 failed to node2; does storage1 failed to storage2?

  2. If storage1 attached to node1, and storage1 failed; does storage1 failed to storage2?

  3. How we view the rules of failed over.?

Please help!!! Thank you so much in advance.

thanks
Dunstan:confused:

2

the documentation is here:
Solaris Cluster 3.2 - Sun Microsystems

and here are some nice basics:
BigAdmin Feature Article: Installing and Configuring Sun Cluster 3.1 09/04 Software for High-Availability Applications

3

Thanks DukeNuke2,

How storage in general is seen within the cluster? This is the portion that i still unclear.

Did it seen as one logical disk; regardless on the 2 Physical (storage1 and storage2)?

Thank you. Dunstan

4

this depends on the storage and how it is configured. there isn't a single answer for that. if you don't know storage basics in solaris, sun cluster might be a little ahead for you...

5

simka, you assumptions are wrong. Both nodes in a cluster see the same (hopefully redundant) filesystem so there is no "storage1 fail" situation. The nodes can fail but the storage isn't supposed to suffer failures.

6

this is not (always) true. this depends also on the configuration and the used HA service. but in the most common configurations you are right.

7

I never saw such a configuration but you are correct. In these uncommon cases where storage isn't shared, I would assume there is no persistence requirements. That would also imply a local storage failure trigger a node switch just like anything else that fails the clustered service.

8

often a geocluster is a good example for a local only storage...

9

I understand geocluster to require data replication using one of AVS, TrueCopy or RDF.

10

Thanks guys, I do have some knowledge on the AIX's LVM work. Eventually, did the solaris behave in such so.

When there are 2 storages (Storage1 and Storage2). On cluster point of view, only LV can be seem; these regardless on which Pyshical Storage (Storage1 and storage2) having issue. Is that concept correct?

------------------ -----------------
| Node1(PRI) | |Node2(SEC) |
------------------ ----------------
/ /
v v
--------------------
|Logical Volume| <----- Cluster concern
-------------------
--------------------------------------------------------------------------
------------------ -----------------
| storage1 | mirror |Storage2 | <---- Correct?
------------------ ----------------

11

In your drawing (please add code tags for it to be readable), there is a single mirrored storage.