This site best when viewed with a modern standards-compliant browser. We recommend Firefox Get Firefox!.

Linux-HA project logo
Providing Open Source High-Availability Software for Linux and other OSes since 1999.

USA Flag UK Flag

Japanese Flag







This web page is no longer maintained. Information presented here exists only to avoid breaking historical links.
The Project stays maintained, and lives on: see the Linux-HA Reference Documentation.
To get rid of this notice, you may want to browse the old wiki instead.



Heartbeat 2.1.4



OSC2007 Tokyo/Fall で Heartbeat紹介

Last site update:
2022-10-02 17:57:22

IBM ServeRAID SCSI RAID Controllers

Certain of the IBM ServeRAID SCSI RAID controllers provide support for what the ServeRAID people refer to as "clustering". Their clustering support provides access to a given RAID volume by one of two machines at a time. That is, two ServeRAID controllers can be connected to the same SCSI array, and then either (but not both) of the two machines can access the data. This is actually perfect for failover, as the firmware guarantees that only one machine at a time can have access to the data, eliminating the need for a STONITH device. This makes a ServeRAID resource a SelfFencingResource. This is a good trait for a failover cluster, and a bad one for many parallel clusters like a scientific cluster. It does work with DB2 parallel clustering, but not with Oracle Parallel server (i.e., RAC -- Real Application Cluster). Oracle in failover mode, yes, but OPS/RAC, no.

Note that using DRBD to mirror a ServeRAID volume does not require clustering support.

The following ServeRAID cards are known to work in clustering mode (and therefore with Linux-HA)

Linux-HA supports all the following ServeRAID cards with its ServeRAID resource agent

The so-called "zero-slot" cards (ServeRAID 5*, or ServeRAID 6i) are not supported. Although there is reason to think that some of the older cards (ServeRAID 3, II, ...) might work, we have not tested them, and they're quite slow by modern standards.

For a listing of these controllers along with pictures and their features, please see the ServeRAID adapter quick reference guide. Note that this guide also indicates which ServeRAID adapters support clustering.

When you use the ServeRAID adapters make sure you have the latest firmware and BIOS loaded onto the adapters, that you have the latest driver and latest ipssend utility. Make special note of the fact that the driver and firmware versions must match, or BadThingsWillHappen. Additionally, the firmware must be version 6.10 or later.

You can find the latest release of the ServeRAID software at IBM Support - ServeRAID Software matrix.

The supported RAID levels for failover are all that have a 1 in them (1, 10, 1E, ...). Any sort of RAID 5 level is NOT supported due to the fact that the firmware does not properly handle certain error recovery cases.

RenzoAlejandroGranados has a Linux-HA + ServeRAID HOWTO which I'll try and get him to move here ;-)

Trademark Notice

ServeRAID is a trademark of IBM.

ServeRAID Clustering

As noted above, ServeRAID shared volumes are self-fencing. Under normal circumstances this means you get three separate lines of protection against split-brain problems with ServeRAID devices.

  1. Heartbeat ensures that no more than one node owns a resource by using its redundant communication mechanisms.

  2. ServeRAID controllers talk to each other over the SCSI bus to ensure that only one of them has access to a given RAID volume at a time.
  3. ServeRAID controllers use SCSI reserve/release to ensure that only one of them has access to the volumes in a ServeRAIDMergeGroup.

So, you're significantly protected against split-brain problems without the use of a STONITH device. The probability of all of the mechanisms failing is extremely low.

See Also

EXP300, EXP400, Heartbeat ServeRAID resource agent, Configuring ServeRAID Hardware, Software configuration for ServeRAID