* [Cluster-devel] cluster/fence/man fence_scsi.8
@ 2006-07-10 22:38 rohara
0 siblings, 0 replies; 3+ messages in thread
From: rohara @ 2006-07-10 22:38 UTC (permalink / raw)
To: cluster-devel.redhat.com
CVSROOT: /cvs/cluster
Module name: cluster
Changes by: rohara at sourceware.org 2006-07-10 22:38:53
Added files:
fence/man : fence_scsi.8
Log message:
Initial version of the fence_scsi man page.
Patches:
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/fence/man/fence_scsi.8.diff?cvsroot=cluster&r1=NONE&r2=1.1
/cvs/cluster/cluster/fence/man/fence_scsi.8,v --> standard output
revision 1.1
--- cluster/fence/man/fence_scsi.8
+++ - 2006-07-10 22:38:53.507621000 +0000
@@ -0,0 +1,110 @@
+.\" Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
+.\" Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+.\"
+.\" This copyrighted material is made available to anyone wishing to use,
+.\" modify, copy, or redistribute it subject to the terms and conditions
+.\" of the GNU General Public License v.2.
+
+.TH fence_apc 8
+
+.SH NAME
+fence_scsi - I/O fencing agent for SCSI persistent reservations
+
+.SH SYNOPSIS
+.B
+fence_scsi
+[\fIOPTION\fR]...
+
+.SH DESCRIPTION
+fence_scsi is an I/O fencing agent which can be used with the SCSI
+devices that support persistent reservations (SPC-2 or greater).
+
+SCSI persistent reservations work by having each node in the cluster
+register with the SCSI device. Registration is done using a unique key
+(based on the node's IP address). Each node that will perform I/O
+operations to the shared storage must register with the device. This
+is done at system startup with the scsi_reserve init script. This
+script will discover all clustered volumes as well as the underlying
+SCSI device(s). Device discovery is done via the lvs command (see
+lvs(8)) and is subject to any filtering rules defined in the lvm.conf
+file.
+
+After generating the node's unique key, the script will register the
+node with the SCSI device(s) that were discovered. Once the node is
+registered, the script will attempt to create a reservation. Unlike
+registrations, of which there are multiple registrants (one for each
+node in the cluster), there is only one reservation holder. If a
+reservation does not already exist for a device, the script will
+create a reservation using the node's unique key.
+
+It is important to distinguish between registrations and
+reservations. As mentioned above, each node that will perform I/O
+operations to the SCSI device must register. As a result, there will
+be multiple registrations for a given SCSI device. In contrast, there
+can only be one reservation per SCSI device. It is not important which
+node holds the reservation. The reservation simply tells the device
+how the registrants are allowed to access the device. For our
+purposes, the reservation type is "write exclusive, registrants only".
+With this reservation type, only registered nodes will be able to
+write to the device.
+
+When the cluster must fence a node, it simply revokes a node's
+registration, or "unregisters" the node. This operation also uses the
+node's unique key. By deriving the unique key based on the errant
+node's IP address, the cluster can "unregister" the key. As a
+result, the errant node will no longer be able to write to the SCSI
+device.
+
+Note that the node that holds the reservation for a device must also
+be registered with that device. When the situation arises where the
+node that is being fenced is also the reservation holder, the
+reservation must be moved. This is handled by using the
+"preempt-and-abort command" which will transfer the reservation from
+the node that is being fenced to the node that is performing the
+fencing. This operation will maintain the reservation while
+"unregistering" the node being fenced.
+
+At system shutdown, the scsi_reserve script will attempt to
+"unregister" the node from all devices. The exception is when the
+node happens to be the reservation holder. In this case, the script
+does nothing, due to the fact that there may be other nodes using the
+device and the reservation must remain intact.
+
+fence_scsi accepts options on the command line as well as from stdin.
+fenced sends parameters through stdin when it execs the agent. fence_scsi
+can be run by itself with command line options. This is useful for testing
+and for turning outlets on or off from scripts.
+
+.SH OPTIONS
+.TP
+\fB-n\fP \fInode\fR
+Name of the node to be fenced.
+.TP
+\fB-h\fP
+Print out a help message describing available options, then exit.
+.TP
+\fB-s\fP \fIself\fR
+Name of the node that will perform the fencing operation.
+.TP
+\fB-v\fP
+Verbose output.
+.TP
+\fB-V\fP
+Print out a version message, then exit.
+
+.SH STDIN PARAMETERS
+.TP
+\fIagent = < param >\fR
+This option is used by fence_node(8) and is ignored by fence_scsi.
+.TP
+\fInodename = < hostname | ip >\fR
+Name of the node to be fenced.
+.TP
+\fIself = < nodename >\fR
+Name of the node that will perform the fencing operation.
+.TP
+\fIverbose = < param >\fR
+Verbose output.
+
+.SH SEE ALSO
+fence(8), fence_node(8), sg_persist(8), lvs(8), lvm.conf(5)
^ permalink raw reply [flat|nested] 3+ messages in thread
* [Cluster-devel] cluster/fence/man fence_scsi.8
@ 2006-07-11 15:08 rohara
0 siblings, 0 replies; 3+ messages in thread
From: rohara @ 2006-07-11 15:08 UTC (permalink / raw)
To: cluster-devel.redhat.com
CVSROOT: /cvs/cluster
Module name: cluster
Changes by: rohara at sourceware.org 2006-07-11 15:08:06
Modified files:
fence/man : fence_scsi.8
Log message:
Updated copywrite and fixed title.
Patches:
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/fence/man/fence_scsi.8.diff?cvsroot=cluster&r1=1.1&r2=1.2
--- cluster/fence/man/fence_scsi.8 2006/07/10 22:38:53 1.1
+++ cluster/fence/man/fence_scsi.8 2006/07/11 15:08:06 1.2
@@ -1,11 +1,10 @@
-.\" Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
-.\" Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+.\" Copyright (C) 2006 Red Hat, Inc. All rights reserved.
.\"
.\" This copyrighted material is made available to anyone wishing to use,
.\" modify, copy, or redistribute it subject to the terms and conditions
.\" of the GNU General Public License v.2.
-.TH fence_apc 8
+.TH fence_scsi 8
.SH NAME
fence_scsi - I/O fencing agent for SCSI persistent reservations
^ permalink raw reply [flat|nested] 3+ messages in thread
* [Cluster-devel] cluster/fence/man fence_scsi.8
@ 2007-01-15 20:05 rohara
0 siblings, 0 replies; 3+ messages in thread
From: rohara @ 2007-01-15 20:05 UTC (permalink / raw)
To: cluster-devel.redhat.com
CVSROOT: /cvs/cluster
Module name: cluster
Branch: RHEL4
Changes by: rohara at sourceware.org 2007-01-15 20:04:59
Added files:
fence/man : fence_scsi.8
Log message:
fence_scsi man page
Patches:
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/fence/man/fence_scsi.8.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=NONE&r2=1.2.6.1
/cvs/cluster/cluster/fence/man/fence_scsi.8,v --> standard output
revision 1.2.6.1
--- cluster/fence/man/fence_scsi.8
+++ - 2007-01-15 20:05:00.028388000 +0000
@@ -0,0 +1,109 @@
+.\" Copyright (C) 2006 Red Hat, Inc. All rights reserved.
+.\"
+.\" This copyrighted material is made available to anyone wishing to use,
+.\" modify, copy, or redistribute it subject to the terms and conditions
+.\" of the GNU General Public License v.2.
+
+.TH fence_scsi 8
+
+.SH NAME
+fence_scsi - I/O fencing agent for SCSI persistent reservations
+
+.SH SYNOPSIS
+.B
+fence_scsi
+[\fIOPTION\fR]...
+
+.SH DESCRIPTION
+fence_scsi is an I/O fencing agent which can be used with the SCSI
+devices that support persistent reservations (SPC-2 or greater).
+
+SCSI persistent reservations work by having each node in the cluster
+register with the SCSI device. Registration is done using a unique key
+(based on the node's IP address). Each node that will perform I/O
+operations to the shared storage must register with the device. This
+is done at system startup with the scsi_reserve init script. This
+script will discover all clustered volumes as well as the underlying
+SCSI device(s). Device discovery is done via the lvs command (see
+lvs(8)) and is subject to any filtering rules defined in the lvm.conf
+file.
+
+After generating the node's unique key, the script will register the
+node with the SCSI device(s) that were discovered. Once the node is
+registered, the script will attempt to create a reservation. Unlike
+registrations, of which there are multiple registrants (one for each
+node in the cluster), there is only one reservation holder. If a
+reservation does not already exist for a device, the script will
+create a reservation using the node's unique key.
+
+It is important to distinguish between registrations and
+reservations. As mentioned above, each node that will perform I/O
+operations to the SCSI device must register. As a result, there will
+be multiple registrations for a given SCSI device. In contrast, there
+can only be one reservation per SCSI device. It is not important which
+node holds the reservation. The reservation simply tells the device
+how the registrants are allowed to access the device. For our
+purposes, the reservation type is "write exclusive, registrants only".
+With this reservation type, only registered nodes will be able to
+write to the device.
+
+When the cluster must fence a node, it simply revokes a node's
+registration, or "unregisters" the node. This operation also uses the
+node's unique key. By deriving the unique key based on the errant
+node's IP address, the cluster can "unregister" the key. As a
+result, the errant node will no longer be able to write to the SCSI
+device.
+
+Note that the node that holds the reservation for a device must also
+be registered with that device. When the situation arises where the
+node that is being fenced is also the reservation holder, the
+reservation must be moved. This is handled by using the
+"preempt-and-abort command" which will transfer the reservation from
+the node that is being fenced to the node that is performing the
+fencing. This operation will maintain the reservation while
+"unregistering" the node being fenced.
+
+At system shutdown, the scsi_reserve script will attempt to
+"unregister" the node from all devices. The exception is when the
+node happens to be the reservation holder. In this case, the script
+does nothing, due to the fact that there may be other nodes using the
+device and the reservation must remain intact.
+
+fence_scsi accepts options on the command line as well as from stdin.
+fenced sends parameters through stdin when it execs the agent. fence_scsi
+can be run by itself with command line options. This is useful for testing
+and for turning outlets on or off from scripts.
+
+.SH OPTIONS
+.TP
+\fB-n\fP \fInode\fR
+Name of the node to be fenced.
+.TP
+\fB-h\fP
+Print out a help message describing available options, then exit.
+.TP
+\fB-s\fP \fIself\fR
+Name of the node that will perform the fencing operation.
+.TP
+\fB-v\fP
+Verbose output.
+.TP
+\fB-V\fP
+Print out a version message, then exit.
+
+.SH STDIN PARAMETERS
+.TP
+\fIagent = < param >\fR
+This option is used by fence_node(8) and is ignored by fence_scsi.
+.TP
+\fInodename = < hostname | ip >\fR
+Name of the node to be fenced.
+.TP
+\fIself = < nodename >\fR
+Name of the node that will perform the fencing operation.
+.TP
+\fIverbose = < param >\fR
+Verbose output.
+
+.SH SEE ALSO
+fence(8), fence_node(8), sg_persist(8), lvs(8), lvm.conf(5)
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2007-01-15 20:05 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-07-10 22:38 [Cluster-devel] cluster/fence/man fence_scsi.8 rohara
-- strict thread matches above, loose matches on Subject: below --
2006-07-11 15:08 rohara
2007-01-15 20:05 rohara
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).