Viewing: obdfilter.failure_domain.4
.TH OBDFILTER...FAILURE_DOMAIN 4 2026-04-07 Lustre "Lustre Kernel Interfaces"
.SH NAME
obdfilter.failure_domain \- control the failure_domain for OSTs
.SH SYNOPSIS
.SY
.BI "lctl get_param obdfilter." FSNAME "-OST*.failure_domain"
.SY
.BI "lctl set_param obdfilter." FSNAME "-OST*.failure_domain=" DOMAIN_NUMBER
.YS
.SS PROPERTIES
.TP
.B Access Permissions
.br
.BR 644 " | " -rw-r--r--
.TP
.B Scope
.br
Per-OST device.
.TP
.B Config
.br
Always present on OST devices.
.TP
.B Default
.br
.RB "obdfilter." FSNAME "-OST*.failure_domain="
.br
.TP
.B Valid Range
.br
.RB "obdfilter." FSNAME "-OST*.failure_domain=0"
.br
.RB "obdfilter." FSNAME "-OST*.failure_domain=4294967295"
.SH DESCRIPTION
The integer
.I DOMAIN_NUMBER
specifies the failure-domain for the OST.
OSTs that share some (external to Lustre) resources such as storage enclosures,
network switches, power supplies, or similar should have different
.I DOMAIN_NUMBER
values (within the constraints of the available storage hardware),
so that file layouts which need to store data redundantly (FLR EC or mirrors)
can allocate objects on independent failure domains.
There will typically be multiple OSTs that share a single
.IR DOMAIN_NUMBER ,
so that multiple copies of redundant file data are not stored on OSTs that
may become unavailable at the same time.
No specific meaning is attributed to the numerical value of
.I DOMAIN_NUMBER
assigned to each OST (e.g. server, rack, row, building),
other than whether they are the same or different.
Except for the value 0. This value means no failure-domain is configured and
failure-domains should be ignored for this OST.
When allocating files using erasure-coding (--ec) the coverage for
failure-domain uniqueness differs between normal allocations (-c) and
allocations using overstriping (-C).
When normal allocation (-c) is used all stripes in the data and its
corresponding ec component must all be in different failure-domains.
When overstriping (-C) is used it is sufficient that all data and ec stripes
within each raid set are from different failure-domains but the same
failure-domain can be used from different raid sets.
When used it is recommended that failure-domain is configured for all OSTs.
For EC protected files it is recommended to use overstriping (-C) as all
raid sets are protected individually.
Currently this is only available for EC protected files.
.SH MODULES
This parameter is in the following modules:
.EX
.RB "obdfilter." FSNAME "-OST*.failure_domain"
.EE
.SH EXAMPLES
.PP
Set the failure domain to 1 for an OST:
.EX
.RB "oss# " "lctl set_param obdfilter.lustrewt-OST0000.failure_domain=1"
.PP
Get the failure domain of an OST:
.EX
.RB "oss# " "lctl get_param obdfilter.lustrewt-OST0000.failure_domain"
1
.EE
.SH AVAILABILITY
.B obdfilter.*.failure_domain
is part of the
.BR lustre (7)
filesystem package since release 2.18.0.
.\" Added in 2.17.51.xxx
.SH SEE ALSO
.BR lctl-get_param (8),
.BR lctl-set_param (8)
