|
Hi folks,
I am getting close to ordering components for a High Availability web server
that will use two Linux servers, each of which is attached to a single
external RAID 5 box using two external RAID controllers in an active-active
configuration. My question is whether anyone has any knowledge or
experience with having two different machines each mount one of two logical
partitions on a single RAID 5 device.
The configuration that I am considering uses two server boxes, each of which
will have two ethernet interfaces and a SCSI adaptor (probably Adaptec
2940U2W). Each server will run a subset of our business software,
specifically one machine will run an apache web server while the other one
will run the backend mysql daemon. The goal is to reduce the single points
of failure and make every component redundant; reliability is more important
than performance for our needs. There will be an external RAID box with two
controllers (probably Mylex DAC960SXI) in an active-active configuration in
which we will build a RAID 5 device. The RAID 5 diskset will be divided
into two logical partitions, and each machine will mount one of those
partitions; one partition will hold the web-specific data and apache
configuration, the other will hold the databases and mysql configuration.
In the event of a failure, I plan to use heartbeat to orchestrate:
- the unmounting of the external partition on the failed server
- shutdown of the failed server
- take over of the failed server's IP and ethernet address by the remaining
good server
- mounting the now idle external partition by the remaining server
- starting the services that were previously running on the failed server.
Does anyone see problems with dividing a RAID 5 diskset into two logical
partitions and mounting each partition by a different machine as long as
only one machine mounts and accesses those partitions at the same time?
Does anyone have general comments regarding this architecture?
-Ben Humphrey
|
