Hi again,
attached is a bit more polished revision.
Best,
Vladislav
22.02.2011 15:01, Vladislav Bogdanov wrote:
> Hi Dejan,
>
> 22.02.2011 13:02, Dejan Muhamedagic wrote:
>> Hi,
>>
>> Where can you get STP stuff from? How to interpret it? And then
>
> Please look at attached RA.
>
> I decided that today is a good time to finally brace myself to find 5
> hours to write it, thanks Frederik ;) .
> Tested, "works for me" (c) in that configuration I talked earlier - STP
> bridge over 1x10Gbps eth + 2x1Gpbs bond.
> (Hopefully) Supports any combination of bridges, bonds, vlans and
> physical ethernet interfaces.
> Tries to guess correct upstream bridge ports, can't test it more
> thoroughly due to absence of more switch hardware (I currently have only
> one c3570x stack per cluster).
> May require some additional checks to be included.
> Also it is linux-specific and requires bash because of my laziness and
> fact that I mainly use Fedora which has nothing against bash yet.
> Can be considered for inclusion in resource-agents (with common license,
> GPLv2?).
>
>> how do you know it's something that won't change within next five
>> minutes? Finally, every failover can incur downtime, is it worth
>> the trouble because what you want is just more performance?
>
> This could be controlled by non-inf location score and f.e. time-based
> stickiness.
> Anyways, I'd better have 10 seconds lockup rather than 10Mb/s per-client
> read for long time when second cluster node (32 disks in HW RAID10) is
> able to easily give another 250-400Mb/s of aggregate throughput.
>
>> Perhaps you don't even need the extra performance at the time.
>
> This depends on what SLA I provide services with...
>
>> Other than that it sounds interesting :-)
>
> Then, please look at the implementation ;)
>
> Best,
> Vladislav
>
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker@oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs:
> http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
#!/bin/bash
#
# OCF resource agent which monitors state of network interface and records it
as a value in CIB
# based on summ of speeds of its active underlying interfaces.
#
# Copyright (c) 2011 Vladislav Bogdanov <bub...@hoster-ok.com>
# Partially based on 'ping' RA by Andrew Beekhof
#
# OCF instance parameters:
# OCF_RESKEY_name: name of attribute to set in CIB
# OCF_RESKEY_iface: network interface to monitor
# OCF_RESKEY_bridge_ports: if not null and OCF_RESKEY_iface is a bridge,
list of bridge ports to consider.
# Default is all ports which have
designated_bridge=root_id
# OCF_RESKEY_weight_base: weight of each 10Mbps in interface speed (1Gbps =
100 * 10Mbps) in CIB score points
#
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/resource.d/heartbeat}
. ${OCF_FUNCTIONS_DIR}/.ocf-shellfuncs
# Defaults
OCF_RESKEY_name_default="ifstatus"
OCF_RESKEY_bridge_ports_default="detect"
OCF_RESKEY_weight_base_default=10
OCF_RESKEY_dampen_default=5
: ${OCF_RESKEY_name=${OCF_RESKEY_name_default}}
: ${OCF_RESKEY_bridge_ports=${OCF_RESKEY_bridge_ports_default}}
: ${OCF_RESKEY_weight_base=${OCF_RESKEY_weight_base_default}}
: ${OCF_RESKEY_dampen=${OCF_RESKEY_dampen_default}}
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="ping">
<version>1.0</version>
<longdesc lang="en">
Every time the monitor action is run, this resource agent records (in the CIB)
speed of active network interfaces from a list.
</longdesc>
<shortdesc lang="en">Network interface status</shortdesc>
<parameters>
<parameter name="name" unique="1">
<longdesc lang="en">
The name of the attributes to set. This is the name to be used in the
constraints.
</longdesc>
<shortdesc lang="en">Attribute name</shortdesc>
<content type="string" default="${OCF_RESKEY_name_default}"/>
</parameter>
<parameter name="iface" unique="0" required="1">
<longdesc lang="en">
Network interface to monitor.
</longdesc>
<shortdesc lang="en">Network interface</shortdesc>
<content type="string" default=""/>
</parameter>
<parameter name="bridge_ports" unique="0">
<longdesc lang="en">
If not null and OCF_RESKEY_iface is a bridge, list of bridge ports to consider.
Default is all ports which have designated_bridge=root_id.
</longdesc>
<shortdesc lang="en">Bridge ports</shortdesc>
<content type="string" default="${OCF_RESKEY_bridge_ports_default}"/>
</parameter>
<parameter name="weight_base" unique="0">
<longdesc lang="en">
Weight of each 10Mbps in interface speed (1Gbps = 100 * 10Mbps).
With default value 1Gbps interface will be counted as 1000.
</longdesc>
<shortdesc lang="en">Weight of 10Mbps interface</shortdesc>
<content type="integer" default="${OCF_RESKEY_weight_base_default}"/>
</parameter>
<parameter name="dampen" unique="0">
<longdesc lang="en">
The time to wait (dampening) further changes occur
</longdesc>
<shortdesc lang="en">Dampening interval</shortdesc>
<content type="integer" default="${OCF_RESKEY_dampen_default}"/>
</parameter>
<parameter name="debug" unique="0">
<longdesc lang="en">
Enables to use default attrd_updater verbose logging on every call.
</longdesc>
<shortdesc lang="en">Verbose logging</shortdesc>
<content type="string" default="false"/>
</parameter>
</parameters>
<actions>
<action name="start" timeout="30" />
<action name="stop" timeout="30" />
<action name="reload" timeout="30" />
<action name="monitor" depth="0" timeout="30" interval="10"/>
<action name="meta-data" timeout="5" />
<action name="validate-all" timeout="30" />
</actions>
</resource-agent>
END
}
ifstatus_usage() {
cat <<END
usage: $0 {start|stop|monitor|migrate_to|migrate_from|validate-all|meta-data}
Expects to have a fully populated OCF RA-compliant environment set.
END
}
ifstatus_start() {
ifstatus_monitor
if [ $? = $OCF_SUCCESS ]; then
return $OCF_SUCCESS
fi
touch ${OCF_RESKEY_pidfile}
ifstatus_update
}
ifstatus_stop() {
rm -f ${OCF_RESKEY_pidfile}
attrd_updater -D -n $OCF_RESKEY_name -d $OCF_RESKEY_dampen $attrd_options
return $OCF_SUCCESS
}
ifstatus_monitor() {
if [ -f ${OCF_RESKEY_pidfile} ]; then
ifstatus_update
return $OCF_SUCCESS
fi
return $OCF_NOT_RUNNING
}
ifstatus_validate() {
# Is the state directory writable?
state_dir=`dirname "$OCF_RESKEY_pidfile"`
touch "$state_dir/$$"
if [ $? != 0 ]; then
ocf_log err "Invalid location for 'state': $state_dir is not writable"
return $OCF_ERR_ARGS
fi
rm "$state_dir/$$"
# Pidfile better be an absolute path
case $OCF_RESKEY_pidfile in
/*) ;;
*) ocf_log warn "You should use an absolute path for pidfile not:
$OCF_RESKEY_pidfile" ;;
esac
# Check the check interval
if ocf_is_decimal "$OCF_RESKEY_CRM_meta_interval" && [
$OCF_RESKEY_CRM_meta_interval -gt 0 ]; then
:
else
ocf_log err "Invalid check interval $OCF_RESKEY_interval. It should be
positive integer!"
exit $OCF_ERR_CONFIGURED
fi
# Check the intarfaces list
if [ "x" = "x$OCF_RESKEY_iface" ]; then
ocf_log err "Empty iface parameter. Please specify some network
interface to check"
exit $OCF_ERR_CONFIGURED
fi
return $OCF_SUCCESS
}
ifstatus_iface_get_speed() {
local iface=$1
local operstate
local carrier
local speed
if [ ! -e "/sys/class/net/$iface" ] ; then
echo 0
elif ifstatus_iface_is_bridge $iface ; then
ifstatus_bridge_get_speed $iface
elif ifstatus_iface_is_bond $iface ; then
ifstatus_bond_get_speed $iface
elif ifstatus_iface_is_vlan $iface ; then
ifstatus_iface_get_speed $( ifstatus_vlan_get_phy $iface )
else
read operstate < "/sys/class/net/$iface/operstate"
read carrier < "/sys/class/net/$iface/carrier"
if [ "$operstate" != "up" ] || [ "$carrier" != "1" ] ; then
speed="0"
else
read speed < "/sys/class/net/$iface/speed"
fi
echo $speed
fi
}
ifstatus_iface_is_vlan() {
local iface=$1
[ -e "/proc/net/vlan/$iface" ] && return 0 || return 1
}
ifstatus_iface_is_bridge() {
local iface=$1
[ -e "/sys/class/net/$iface/bridge" ] && return 0 || return 1
}
ifstatus_iface_is_bond() {
local iface=$1
[ -e "/sys/class/net/$iface/bonding" ] && return 0 || return 1
}
ifstatus_vlan_get_phy() {
local iface=$1
grep "^$iface " "/proc/net/vlan/config" | sed -r 's/.*\| +(.*)/\1/'
}
ifstatus_bridge_is_stp_enabled() {
local iface=$1
local stp
read stp < "/sys/class/net/$iface/bridge/stp_state"
[ "$stp" = "1" ] && return 0 || return 1
}
ifstatus_bridge_get_root_ports() {
local bridge=$1
local root_id
local root_ports=""
local bridge_id
read root_id < "/sys/class/net/$bridge/bridge/root_id"
for port in /sys/class/net/$bridge/brif/* ; do
read bridge_id < "$port/designated_bridge"
if [ "$bridge_id" = "$root_id" ] ; then
root_ports="$root_ports ${port##*/}"
fi
done
echo "${root_ports# }"
}
# From /inlude/linux/if_bridge.h:
#define BR_STATE_DISABLED 0
#define BR_STATE_LISTENING 1
#define BR_STATE_LEARNING 2
#define BR_STATE_FORWARDING 3
#define BR_STATE_BLOCKING 4
ifstatus_bridge_get_active_ports() {
local bridge=$1
shift 1
local ports="$*"
local active_ports=""
local port_state
local stp_state=ifstatus_bridge_is_stp_enabled $bridge
local warn=0
if [ -z "$ports" ] || [ "$ports" = "detect" ] ; then
ports=$( ifstatus_bridge_get_root_ports $bridge )
fi
for port in $ports ; do
if [ ! -e "/sys/class/net/$bridge/brif/$port" ] ; then
ocf_log warning "Port $port doesn't belong to bridge $bridge"
continue
fi
read port_state < "/sys/class/net/$bridge/brif/$port/state"
if [ "$port_state" = "3" ] ; then
if [ -n "$active_ports" ] && $stp_state ; then
warn=1
fi
active_ports="$active_ports $port"
fi
done
if [ $warn -eq 1 ] ; then
ocf_log warning "More then one upstream port in bridge '$bridge' is in
forwarding state while STP is enabled: $active_ports"
fi
echo "${active_ports# }"
}
ifstatus_bridge_get_speed() {
local $iface=$1
if ! ifstatus_iface_is_bridge $iface ; then
echo 0
return
fi
local ports=$( ifstatus_bridge_get_active_ports $iface
${OCF_RESKEY_bridge_ports} )
for port in $ports ; do
: $(( aggregate_speed += $( ifstatus_iface_get_speed $port ) ))
done
echo $aggregate_speed
}
ifstatus_bond_get_slaves() {
local iface=$1
local slaves
read slaves < "/sys/class/net/$iface/bonding/slaves"
echo $slaves
}
ifstatus_bond_get_active_iface() {
local iface=$1
local active
read active < "/sys/class/net/$iface/bonding/active_slave"
echo $active
}
ifstatus_bond_is_balancing() {
local iface=$1
read mode mode_index < "/sys/class/net/$iface/bonding/mode"
case $mode in
"balance-rr"|"balance-xor"|"802.3ad"|"balance-tlb"|"balance-alb")
return 0
;;
*)
return 1
;;
esac
}
ifstatus_bond_get_speed() {
local iface=$1
local aggregate_speed=0
if ! ifstatus_iface_is_bond $iface ; then
echo 0
return
fi
local slaves=$( ifstatus_bond_get_slaves $iface )
if ifstatus_bond_is_balancing $iface ; then
for slave in $slaves ; do
: $(( aggregate_speed += $( ifstatus_iface_get_speed $slave ) ))
done
# Bonding is unable to get speed*n
: $(( aggregate_speed = aggregate_speed*8/10 ))
else
: $(( aggregate_speed = $( ifstatus_iface_get_speed $(
ifstatus_bond_get_active_iface $iface ) ) ))
fi
echo $aggregate_speed
}
ifstatus_update() {
local speed=$( ifstatus_iface_get_speed $OCF_RESKEY_iface)
: $(( score = speed * $OCF_RESKEY_weight_base / 10 ))
attrd_updater -n $OCF_RESKEY_name -v $score -d $OCF_RESKEY_dampen
$attrd_options
rc=$?
case $rc in
0)
ocf_is_true ${OCF_RESKEY_debug} && ocf_log debug "Updated
$OCF_RESKEY_name = $score"
;;
*)
ocf_log warn "Could not update $OCF_RESKEY_name = $score: rc=$rc"
;;
esac
return $rc
}
if [ `uname` != "Linux" ] ; then
ocf_log err "This RA works only on linux."
exit $OCF_ERR_INSTALLED
fi
if ! ocf_is_true ${OCF_RESKEY_CRM_meta_globally_unique} ; then
: ${OCF_RESKEY_pidfile:="$HA_VARRUN/ifstatus-${OCF_RESKEY_name}"}
else
: ${OCF_RESKEY_pidfile:="$HA_VARRUN/ifstatus-${OCF_RESOURCE_INSTANCE}"}
fi
attrd_options='-q'
if ocf_is_true ${OCF_RESKEY_debug} ; then
attrd_options=''
fi
case $__OCF_ACTION in
meta-data)
meta_data
exit $OCF_SUCCESS
;;
start)
ifstatus_start
;;
stop)
ifstatus_stop
;;
monitor)
ifstatus_monitor
;;
reload)
ifstatus_start
;;
validate-all)
ifstatus_validate
;;
usage|help)
ifstatus_usage
exit $OCF_SUCCESS
;;
*)
ifstatus_usage
exit $OCF_ERR_UNIMPLEMENTED
;;
esac
exit $?
_______________________________________________
Pacemaker mailing list: Pacemaker@oss.clusterlabs.org
http://oss.clusterlabs.org/mailman/listinfo/pacemaker
Project Home: http://www.clusterlabs.org
Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
