The default for calculating summary route's cost is RFC 1583.
RFC 2328 uses a different method .
Cisco conforms to both but RFC 1583 is default.
So to use RFC 2328 instead
specify following
router ospf 1
no compatible rfc1583
Showing posts with label OSPF. Show all posts
Showing posts with label OSPF. Show all posts
Wednesday, January 06, 2010
Reduce OSPF LSA Flooding v/s Prevent OSPF LSA Flooding.
Earlier I mentioned how to prevent LSA flooding using interface level cmd ip ospf database-filter all out.
Now supposedly, you want to reduce flooding instead of preventing it :-) (Symantecs of Stupid English)
you would apply this cmd
ip ospf flood-reduction
Now supposedly, you want to reduce flooding instead of preventing it :-) (Symantecs of Stupid English)
you would apply this cmd
ip ospf flood-reduction
Jumbo Frames and MTU and OSPF neighbor ship
The moment one changes mtu on ONLY one end of the link to support Jumbo Frames ( > 9000 bytes) , when OSPF is restarted it will fail to form adjacency since MTU mismatch will occur.
To ignore MTU mismatch, specify ip ospf mtu-ignore on both ends of the link.
To ignore MTU mismatch, specify ip ospf mtu-ignore on both ends of the link.
Prevent Flooding of OSPF LSA Broadcasts
Trackback
int s0/0.100
ip ospf database-filter all out
Will allow receiving Hello and LSA from peer
But wont sent out our LSA to them. Our Hellos are still sent out.
==> WE form adjacency with peer. WE get their LSA and install in our routing table
==>Peer forms adjacency with us. peer doesnt get our LSA and hence doesnt see our routes.
==> One way Traffic Flow.... Can be useful in Redundant links / FR Hub n Spoke Topology
The Biggest difference between passive-interface and database-filter is with passive-interface, even hellos are not sent out. So no adjacency and no one way communication.
int s0/0.100
ip ospf database-filter all out
Will allow receiving Hello and LSA from peer
But wont sent out our LSA to them. Our Hellos are still sent out.
==> WE form adjacency with peer. WE get their LSA and install in our routing table
==>Peer forms adjacency with us. peer doesnt get our LSA and hence doesnt see our routes.
==> One way Traffic Flow.... Can be useful in Redundant links / FR Hub n Spoke Topology
The Biggest difference between passive-interface and database-filter is with passive-interface, even hellos are not sent out. So no adjacency and no one way communication.
ignore lsa mospf ==> Do not generate syslog for Type 6 (MOSPF) packets
trackback
And also on CCO
Cisco routers do not support LSA Type 6 MOSPF packets, and they generate syslog messages if they receive such packets. If the router is receiving many MOSPF packets, you might want to configure the router to ignore the packets and thus prevent a large number of syslog messages.
Remember Type 6 LSA = MOSPF LSA
And also on CCO
Cisco routers do not support LSA Type 6 MOSPF packets, and they generate syslog messages if they receive such packets. If the router is receiving many MOSPF packets, you might want to configure the router to ignore the packets and thus prevent a large number of syslog messages.
Remember Type 6 LSA = MOSPF LSA
OSPF LSA flooding pacing
The CCO link explains
ASBR1(config-router)#timers pacing ?
flood OSPF flood pacing timer
lsa-group OSPF LSA group pacing timer
retransmission OSPF retransmission pacing timer
ASBR1(config-router)#timers pacing ?
flood OSPF flood pacing timer
lsa-group OSPF LSA group pacing timer
retransmission OSPF retransmission pacing timer
Feature Overview
In rare situations, you might need to change Open Shortest Path First (OSPF) packet-pacing default timers to mitigate CPU or buffer utilization issues associated with flooding very large numbers of link-state advertisements (LSAs). The OSPF Update Packet-Pacing Configurable Timers feature allows you to configure the rate at which OSPF LSA flood pacing, retransmission pacing, and group pacing updates occur.
Configuring OSPF flood pacing timers allows you to control interpacket spacing between consecutive link-state update packets in the OSPF transmission queue. Configuring OSPF retransmission pacing timers allows you to control interpacket spacing between consecutive link-state update packets in the OSPF retransmission queue. Cisco IOS software groups the periodic refresh of LSAs to improve the LSA packing density for the refreshes in large topologies. The group timer controls the interval used for group LSA refreshment; however, this timer does not change the frequency that individual LSAs are refreshed (the default refresh occurs every 30 minutes).
The default are
LSA group pacing timer 240 secs ==> timers pacing lsa-group
Interface flood pacing timer 33 msecs ==> timers pacing flood
Retransmission pacing timer 66 msecs ==> timers pacing retransmission
OSPF : Configure with Highest Level Security !
=> Must use md5 authentication instead of the three types available ( null, normal,md5)
i.e
router ospf 1
area 0 authentication message-digest
int f0/0
ip ospf message-digest 1 md5 CISCO
OR
int f0/0
ip ospf authentication message-digest
ip ospf message-digest 1 md5 CISCO
i.e
router ospf 1
area 0 authentication message-digest
int f0/0
ip ospf message-digest 1 md5 CISCO
OR
int f0/0
ip ospf authentication message-digest
ip ospf message-digest 1 md5 CISCO
Monday, December 28, 2009
OSPF sham-link
Rules:
1) Can belong to any area including area 0
2) Sham Link end points cannot be advertised by OSPF
steps on PE:
1) Create a New Loopback and put it in vrf
2) Advertise the loopback in MPBGP
3) Create a sham-link
4) Ensure the OPSF cost on the backdoor link is high so that traffic prefers MPLS link.
1) Can belong to any area including area 0
2) Sham Link end points cannot be advertised by OSPF
steps on PE:
1) Create a New Loopback and put it in vrf
2) Advertise the loopback in MPBGP
3) Create a sham-link
4) Ensure the OPSF cost on the backdoor link is high so that traffic prefers MPLS link.
Sunday, April 06, 2008
Day3: OSPF
Day 3
OSPF
Process ID is LOCALLY Significant
Must have atleast one UP/UP inerface trunning IPV4 for router OSPF to pick it as Router-ID
Network Statement : Misconception: Network Statement relates to Subnet Mask of the network that you are trying to oroiginate
network Statement only is used to pick interface that matches the network statement
For e.g : network 0.0.0.0 255.255.255.255 area 0 ==> ALl interfaces , does not mean that we are originating the Default nwk
In older IOS, the order of entering the network statements were significant, NOt anymore, the newer IOS reorders the network statement to order more specific match first
Forming Adjacencies
########Must Match ####
-area
-hello/dead timers
- MTU
-(OR) ip ospf mtu-ignore
-compatible network types
-stub flags
-authentication
#########Must be unique #########
-Intf IP Address
- OSPF Router-ID
When doing OSPF Adj, chk sh system mtu, it must match one witht he router or use "ip mtu" cmd to change the MTU on rtr on certain interfaces
RECOMMEND: Manually define router-ID, assuming addressing is globally unique, but if u use Anycast IP, then u can have IP duplicates
RECOMMEND: Avoid using Rtr-ID 1.1.1.1 or 2.2.2.2 or like, cos BBRs (backbone routers )cud be using the same, and this can casue inconsistencies in the OSPF Databases.
###############Types###############################
-Broadcast
-Non-Broadcast
-Point-to-Point
-Point-to-Multipoint
-Point-to-Multipoint Non-Boradcast
-Loopback
-Broadcast
-Send Hellos as multicast
-224.0.0.5 (all ospf rtrs in seg ) /224.0.0.6 (all DR/BDR rtrs in the seg)
-uses DR/BDR, to minimise LSA Replication 9Like BGP Router Reflector)
in FR Hub and spoke env
- interface level cmd: "ip ospf network broadcast" on all neighbor's interfaces required.
- ip ospf prio 0 req for spoke sites, so as to avoid them becoming DR/BDR
- no neighbor cmds required on hub
- frame-realy map stmts need broadcast keyword
- If you notice: Next HOp Modification doesn't happen.
- if you notice: DR/BDR are elected.
DR/BDR Election
-Priority, 0 ==> wont participate in election, 255 is max and its best
-No premption in DR/BDR election
==> once DR/BDR are elected, later on a higher prio rtr cannot preempt the DR
-Router-ID
if all teh devices in the segment have the same priority, thena the router0-d wil be chosen
Higer rtr-id is better
Rtr-ID doesnt need to be a valid IP or valid routeable Number, its can be any number in dotted decimal format, for e.g 255.255.255.255 will be the most preferred rtr id
-- Caveat: DR/BDR also depends on whiuch Router is configured first and whose OSPF process converges faster.
#########Debug OSPF####
access-list 100 permit 89 any any
debug ip packet 100
How to find who is the DR in Broadcast/NBMA segments using sho ip ospf Database cmd ??????????
Ans:--The "Net Link States" table shows Who is the DR (BDR cant be found) in teh Adv Router Column for the given Segment as represented by the Link ID
Network Type : NBMA (Updates sent as Replicated Unicast) Has DR/BDR Election, but needs neighbor stmts to be added for Updates to be sent to it as unicast
Hello = 30s , Dead = 120s
FR Main Int
FR multipoint sub-int
ATM interfaces
in FR Hub and spoke env
- Default Network Type for FR (NBMA)
- neighbor cmds under ospf process required on hub, since updates sent as Unicast
- frame-realy map stmts DON'T need broadcast keyword
- ip ospf prio 0 req for spoke sites, so as to avoid them becoming DR/BDR
- If you notice: Next HOp Modification doesn't happen.
- if you notice: DR/BDR are elected.
#################OSPF Network Point to Multipoint #####################
-Not a default Option , need to ue intf cmd "ip ospf network point-to-multipoint"
-Sends hellos as Multicast on 224.0.0.5
-Modifies the next hop
-No DR/BDR elecetion
-Adds in routing table Host /32 route for end points of the links, then uses recursive looksups to find the interface.
-Ensure that you have a broadcast keyword in frame-relay map cmd to allow multicast traffic
-U need to have ONLY 1 Mapping from Spoke to HUB. Rest of L2 mapping is taken care at L3 via OSPF mulitpoint technology
The functional difference between OSPF network type point-to-multipoint versus
broadcast and non-broadcast network types is how point-to-multipoint deals with nexthop
resolution on a non-broadcast media. OSPF network type point-to-multipoint treats
the network as a collection of point-to-point links instead of one flat broadcast network.
With the network types non-broadcast and broadcast, OSPF does not understand that
the underlying layer 2 topology may not mirror a flat layer 3 network. With OSPF network
types broadcast and non-broadcast, next hop values are not modified when updates are
transmitted across an NBMA media. This implies that devices on the NBMA cloud
require layer 3 to layer 2 resolution for any endpoint injecting routes into the network.
With OSPF network type point-to-multipoint, next hop values are modified to the address
of the directly connected neighbor when they are advertised across the NBMA cloud.
This implies that routers on the NBMA network only need layer 3 to layer 2 resolution for
directly connected neighbors when running OSPF network point-to-mulitpoint.
in FR Hub and spoke env
- NOT a Default Network Type for FR
- NO neighbor cmds under ospf process required on hub, since updates sent as multicast
- frame-realy map stmts need broadcast keyword, but only "ONE" map stmt to the HUB, rest are resolved by
- ip ospf prio 0 NOT req for spoke sites, NO DR/BDR in this TYpe
- If you notice: Next HOp Modification DOES happen. There are /32 Routes added for the Link address and the Remtoe networks are pointed to this Link /32 Address.
- if you notice: DR/BDR are NOT elected ( Not supported).
#################OSPF Network Point to Multipoint NonBroadcast #######
-Not a Default Option, use intf cmd "ip ospf network point-to-multipoint non-broadcast"
- Sends hellos as unicast
-neighbor stmts required
-no DR/BDR electioin
- Modifies next-hop
-Can be used when we have PVCs with different costs
-in ethernet : diff BW values for diff neighbors in the link
in FR Hub and spoke env
- NOT a Default Network Type for FR
- Neighbor cmds under ospf process required on hub, since updates sent as unicast
- frame-realy map stmts DON'T need broadcast keyword, AND only "ONE" map stmt to the HUB, rest are resolved by L3 (OSPF)
- ip ospf prio 0 NOT req for spoke sites, NO DR/BDR in this TYpe
- If you notice: Next HOp Modification DOES happen. There are /32 Routes added for the Link address and the Remtoe networks are pointed to this Link /32 Address.
- if you notice: DR/BDR are NOT elected ( Not supported).
- if you notice: ONLY this Network Type, allows COST to be modifed under OSPF process when defining neighbors. => neighbor 10.0.0.4 cost 50 , etc
##########How to change the Cost of a Link ? #########################
Typically to be used in scenarios where u have links > 100mbps , for e.g gig, 10g, OC3, etc
- Interface"Bandwidth"
- Interface "ip ospg cost"
- Process "auto-cost"
-Process "neighbor w.x.y.z cost" ==> used in Multipoint
TIP: How to calculate the Cost of an OPSF Link without using the formual ourselves ?
Ans: Configure on of the physcial lkinsk and change their BW in kbps and then type "sh ip ospf interface" to get its cost
############Point to Point #########
-no DR/BDR
-Update sent as Multicast
"ip ospf 1 area 0 " ==> Eabled at Interface level means that we include the interface for a given area, and ALSO, the secondary addresses if any configured on the itnerface are also enabled for OSPF
This means the secondary addresses will also send out "hello" and form adj
in FR Hub and spoke env
- NOT a Default Network Type for FR
- NO Neighbor cmds under ospf process required on hub, since updates sent as multicast
- frame-realy map stmts need broadcast keyword, since only 1 neighbor, only 1 MAP stmt
- ip ospf prio 0 NOT req for spoke sites, NO DR/BDR in this TYpe
- If you notice: Next HOp Modification DOES NOT happen.
- if you notice: DR/BDR are NOT elected ( Not supported).
Summaray: OSPF network Types
Broadcast can be adj to Non-Broadcast as they both have DR/BDR concept, however we need to ensure tha the Hello/Dead timers are matched
Point-to-point can be adj with Point-to-Multipoint OR adj with point-to-multipoint non-broadcast, as they dont have DR/BDR concept, however we need to ensure tha the Hello/Dead timers are matched
V.V IMP: What gets affected with non-compatible nwks but it appears that u are formining adj , but actually its a neighbor relnaship but not truly adj
-unicast/multicast
-DR/BDR Election
-Next-hop processing
######### OSPF fast Hellos ##############
ip ospf dead-interval minimal hello-multiplier 5
- interface level cmd
- dead int = 1s , hello= 200msx5=1000ms =1s
-so hello sent out every 200ms
OSPF
Process ID is LOCALLY Significant
Must have atleast one UP/UP inerface trunning IPV4 for router OSPF to pick it as Router-ID
Network Statement : Misconception: Network Statement relates to Subnet Mask of the network that you are trying to oroiginate
network Statement only is used to pick interface that matches the network statement
For e.g : network 0.0.0.0 255.255.255.255 area 0 ==> ALl interfaces , does not mean that we are originating the Default nwk
In older IOS, the order of entering the network statements were significant, NOt anymore, the newer IOS reorders the network statement to order more specific match first
Forming Adjacencies
########Must Match ####
-area
-hello/dead timers
- MTU
-(OR) ip ospf mtu-ignore
-compatible network types
-stub flags
-authentication
#########Must be unique #########
-Intf IP Address
- OSPF Router-ID
When doing OSPF Adj, chk sh system mtu, it must match one witht he router or use "ip mtu" cmd to change the MTU on rtr on certain interfaces
RECOMMEND: Manually define router-ID, assuming addressing is globally unique, but if u use Anycast IP, then u can have IP duplicates
RECOMMEND: Avoid using Rtr-ID 1.1.1.1 or 2.2.2.2 or like, cos BBRs (backbone routers )cud be using the same, and this can casue inconsistencies in the OSPF Databases.
###############Types###############################
-Broadcast
-Non-Broadcast
-Point-to-Point
-Point-to-Multipoint
-Point-to-Multipoint Non-Boradcast
-Loopback
-Broadcast
-Send Hellos as multicast
-224.0.0.5 (all ospf rtrs in seg ) /224.0.0.6 (all DR/BDR rtrs in the seg)
-uses DR/BDR, to minimise LSA Replication 9Like BGP Router Reflector)
in FR Hub and spoke env
- interface level cmd: "ip ospf network broadcast" on all neighbor's interfaces required.
- ip ospf prio 0 req for spoke sites, so as to avoid them becoming DR/BDR
- no neighbor cmds required on hub
- frame-realy map stmts need broadcast keyword
- If you notice: Next HOp Modification doesn't happen.
- if you notice: DR/BDR are elected.
DR/BDR Election
-Priority, 0 ==> wont participate in election, 255 is max and its best
-No premption in DR/BDR election
==> once DR/BDR are elected, later on a higher prio rtr cannot preempt the DR
-Router-ID
if all teh devices in the segment have the same priority, thena the router0-d wil be chosen
Higer rtr-id is better
Rtr-ID doesnt need to be a valid IP or valid routeable Number, its can be any number in dotted decimal format, for e.g 255.255.255.255 will be the most preferred rtr id
-- Caveat: DR/BDR also depends on whiuch Router is configured first and whose OSPF process converges faster.
#########Debug OSPF####
access-list 100 permit 89 any any
debug ip packet 100
How to find who is the DR in Broadcast/NBMA segments using sho ip ospf Database cmd ??????????
Ans:--The "Net Link States" table shows Who is the DR (BDR cant be found) in teh Adv Router Column for the given Segment as represented by the Link ID
Network Type : NBMA (Updates sent as Replicated Unicast) Has DR/BDR Election, but needs neighbor stmts to be added for Updates to be sent to it as unicast
Hello = 30s , Dead = 120s
FR Main Int
FR multipoint sub-int
ATM interfaces
in FR Hub and spoke env
- Default Network Type for FR (NBMA)
- neighbor cmds under ospf process required on hub, since updates sent as Unicast
- frame-realy map stmts DON'T need broadcast keyword
- ip ospf prio 0 req for spoke sites, so as to avoid them becoming DR/BDR
- If you notice: Next HOp Modification doesn't happen.
- if you notice: DR/BDR are elected.
#################OSPF Network Point to Multipoint #####################
-Not a default Option , need to ue intf cmd "ip ospf network point-to-multipoint"
-Sends hellos as Multicast on 224.0.0.5
-Modifies the next hop
-No DR/BDR elecetion
-Adds in routing table Host /32 route for end points of the links, then uses recursive looksups to find the interface.
-Ensure that you have a broadcast keyword in frame-relay map cmd to allow multicast traffic
-U need to have ONLY 1 Mapping from Spoke to HUB. Rest of L2 mapping is taken care at L3 via OSPF mulitpoint technology
The functional difference between OSPF network type point-to-multipoint versus
broadcast and non-broadcast network types is how point-to-multipoint deals with nexthop
resolution on a non-broadcast media. OSPF network type point-to-multipoint treats
the network as a collection of point-to-point links instead of one flat broadcast network.
With the network types non-broadcast and broadcast, OSPF does not understand that
the underlying layer 2 topology may not mirror a flat layer 3 network. With OSPF network
types broadcast and non-broadcast, next hop values are not modified when updates are
transmitted across an NBMA media. This implies that devices on the NBMA cloud
require layer 3 to layer 2 resolution for any endpoint injecting routes into the network.
With OSPF network type point-to-multipoint, next hop values are modified to the address
of the directly connected neighbor when they are advertised across the NBMA cloud.
This implies that routers on the NBMA network only need layer 3 to layer 2 resolution for
directly connected neighbors when running OSPF network point-to-mulitpoint.
in FR Hub and spoke env
- NOT a Default Network Type for FR
- NO neighbor cmds under ospf process required on hub, since updates sent as multicast
- frame-realy map stmts need broadcast keyword, but only "ONE" map stmt to the HUB, rest are resolved by
- ip ospf prio 0 NOT req for spoke sites, NO DR/BDR in this TYpe
- If you notice: Next HOp Modification DOES happen. There are /32 Routes added for the Link address and the Remtoe networks are pointed to this Link /32 Address.
- if you notice: DR/BDR are NOT elected ( Not supported).
#################OSPF Network Point to Multipoint NonBroadcast #######
-Not a Default Option, use intf cmd "ip ospf network point-to-multipoint non-broadcast"
- Sends hellos as unicast
-neighbor stmts required
-no DR/BDR electioin
- Modifies next-hop
-Can be used when we have PVCs with different costs
-in ethernet : diff BW values for diff neighbors in the link
in FR Hub and spoke env
- NOT a Default Network Type for FR
- Neighbor cmds under ospf process required on hub, since updates sent as unicast
- frame-realy map stmts DON'T need broadcast keyword, AND only "ONE" map stmt to the HUB, rest are resolved by L3 (OSPF)
- ip ospf prio 0 NOT req for spoke sites, NO DR/BDR in this TYpe
- If you notice: Next HOp Modification DOES happen. There are /32 Routes added for the Link address and the Remtoe networks are pointed to this Link /32 Address.
- if you notice: DR/BDR are NOT elected ( Not supported).
- if you notice: ONLY this Network Type, allows COST to be modifed under OSPF process when defining neighbors. => neighbor 10.0.0.4 cost 50 , etc
##########How to change the Cost of a Link ? #########################
Typically to be used in scenarios where u have links > 100mbps , for e.g gig, 10g, OC3, etc
- Interface"Bandwidth"
- Interface "ip ospg cost"
- Process "auto-cost"
-Process "neighbor w.x.y.z cost" ==> used in Multipoint
TIP: How to calculate the Cost of an OPSF Link without using the formual ourselves ?
Ans: Configure on of the physcial lkinsk and change their BW in kbps and then type "sh ip ospf interface" to get its cost
############Point to Point #########
-no DR/BDR
-Update sent as Multicast
"ip ospf 1 area 0 " ==> Eabled at Interface level means that we include the interface for a given area, and ALSO, the secondary addresses if any configured on the itnerface are also enabled for OSPF
This means the secondary addresses will also send out "hello" and form adj
in FR Hub and spoke env
- NOT a Default Network Type for FR
- NO Neighbor cmds under ospf process required on hub, since updates sent as multicast
- frame-realy map stmts need broadcast keyword, since only 1 neighbor, only 1 MAP stmt
- ip ospf prio 0 NOT req for spoke sites, NO DR/BDR in this TYpe
- If you notice: Next HOp Modification DOES NOT happen.
- if you notice: DR/BDR are NOT elected ( Not supported).
Summaray: OSPF network Types
Broadcast can be adj to Non-Broadcast as they both have DR/BDR concept, however we need to ensure tha the Hello/Dead timers are matched
Point-to-point can be adj with Point-to-Multipoint OR adj with point-to-multipoint non-broadcast, as they dont have DR/BDR concept, however we need to ensure tha the Hello/Dead timers are matched
V.V IMP: What gets affected with non-compatible nwks but it appears that u are formining adj , but actually its a neighbor relnaship but not truly adj
-unicast/multicast
-DR/BDR Election
-Next-hop processing
######### OSPF fast Hellos ##############
ip ospf dead-interval minimal hello-multiplier 5
- interface level cmd
- dead int = 1s , hello= 200msx5=1000ms =1s
-so hello sent out every 200ms
Subscribe to:
Posts (Atom)