ospf disconnected area 0

[gavin han]

we have 4 data centers, 4 data center core switches connected to each other using point-to-point trunk link. all these 4 data centers core switches are in area 0.  there are other sites connected to these data centers through edge router at data center and are in area mentioned in the rough paint diagram (attached).

DC-4 lost connectivity connectivity in following 2 cases:

1. all P2P links of DC-4 going to other data center went down so DC-4 became isolated area 0. how to fix it so that we don't loose DC-4 (and users in DC-4 stays up and don't loose connectivity to rest of the network) even if its all P2P links to other DC goes down? as shown in diagram, DC-4 is connected to site-2 through area52, site-2 is also connected to other DC using area52.

2. someone accidentally changed router-id on DC-4 core and configured it same as DC-1 but didn't clear ospf process so change wasn't applicable. DC-4 core reboot and we lost connectivity to it. why did we loose connectivity (users behind DC-4 core were down) to DC-4 since it connects to all other data centers apartment DC-1? how to fix it so that users in DC-4 don't loose connectivity in such cases?

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

#1 was due to DC-4's area zero becoming partitioned.  You need area 0 connections, to each area, for OSPF to pass traffic between areas.

One solution would be to place an area zero router in area 51 and/or area 52.

#2 What you're describing/asking is a bit unclear.  You're saying DC-4 got misconfigured but you don't want any misconfiguration to adversely impact the network?

[gavin han]

#1 I can't configure area0 router in area51 or area52 router. this is example diagram, our network is big so if i put area0 router there then i will have to put alot of routers in area0 and almost all routers will end up in area zero. do you recommend configuring virtual link to other area0 using ABR for area51, area52?

#2 yes,  i don't want configuration adversely affect our network.

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

#1 You would have to explain your topology a bit further.  You don't need to do all remote areas, and depending on your topology, perhaps only one router in the area would need to become an area zero router.  Basically, making an area like 51 or 52, as shown in your diagram, an alternate path for area 0 links, as all the other area zeros DCs are connected.

Yes, virtual links was another solution option I had in mind, as also might be running a logical area 0 link via a GRE tunnel.

#2 Okay, you're still saying you want any misconfiguration to not be adverse to your network?  If so, every hear the old joke "Patient: Doctor, doctor, it hurts when I do this.  Doctor: Then don't do that."

Jokes aside, it's difficult to protect against misconfiguration errors.  Now if you have a single DC that's a potential single point of failure, just as having additional hardware avoids that single point of failure, such additional hardware might protect you from some misconfiguration errors too.

I.e. so regardless of whether DC-4's core router is lost due to a hardware failure or due to misconfiguration, you would need to design/provide redundancy that the rest of the network can tolerate such a lost.

Do understand misconfiguration issues can be harder to deal with than hardware issues, because hardware is more likely to fail outright while a misconfiguration may only partially fail, and such a failure may be difficult to recognize as a failure.

[gavin han]

Thanks for the reply.

#1 more topology description: We have about 200 sites, each site has 2 P2P uplinks connected to one of the data center edge router. each site is in separate areaX. edge router link towards each site is in areaX while its loopback and link to core switch is in area0. distribution switch for DC campus buildings also connects to Core switch, distribution switch link to core switch and its loopback is in area0 while other links and all other vlans are in specific areaY. 

hope this description makes sense :).

what do you mean by "You don't need to do all remote areas, and depending on your topology, perhaps only one router in the area would need to become an area zero router.  Basically, making an area like 51 or 52, as shown in your diagram, an alternate path for area 0 links, as all the other area zeros DCs are connected."

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

what do you mean . . .

From you diagram, it appears that there was only one area 0 link from each DC.  If that were true, losing it would partition that DC's area 0.

If I'm reading your above post correctly, your have area 0 links to each branch, so you already have much of what I was suggesting.

However, if there's only one DC core router per site, and it's the only one with off-site connections, then losing those connections, or the device hosting them, will, of course, partition your site.  So, I'm confused again.  I would imagine you know that, so not sure now what you were asking in your OP.

[gavin han]

NO. remote sites are not in area 0. they are in area X. only DC are in area 0.

[ahmedshoaib]

Hi Gavin;

As I understand that, you have 4 Data Center and each DC have 1 core switch and all DC core switches are connected each other via L2 trunk connection and respective distribution switches. Core switch upstream link connected with edge Router via area 51 & area 52. Please add if I miss any thing to better understand.

Now my question is, will you run ospf between all DC Core Switches (DC1, DC2, DC3, DC4) if yes, then which area you use?

If not I will recommend you to run ospf area 0 between all 4 DC core switches.

Thanks 

Shoaib Ahmed

[paul driver]

Hello

1, Having 4 P2P links between all DC seems to me to have all the resiliency you require, In this case you've experienced a total outage on your Dc interconnects then I guess the answer is even if you had 8 P2P links you cannot account for that.

As for the partitioning, Possibly use a Virtual link between area 52-51, as by default VL's become part of area 0

2) You didn't lose anything until the DC-4 core reloaded as even if you change the RID in an active ospf process, This change will not take affect until either in your case the router is reloaded or the opsf process is restarted.

The router(s) now with the duplicate rids wont re-establish adjacency and the sites would flapp between the DC1 and DC4

res
Paul