BGP Process detected a crash

[loveandrew821]

hello,

Our CISCO 7609 which is using WS-SUP32-GE-3B (ios : 12.2(33)SRD4 ) has detected a crash and automatically reloaded while I was donig 'no address-family' and 'no ip vrf' job. I opend a case and the CISCO Tac gave me an answer that I am hitting a bug 'CSCtg94250'. The bug says as below :

Symptoms: Removing address-family ipv4 vrf (in router BGP) followed by no ip vrf (where "vrf" is the same) could result in a crash.

Conditions: The symptom is observed in a large VPNv4 scale setup, when applyingthe following commands to the same VRF back-to-back:

1. no address-family ipv4 vrf 
2. no ip vrf 
3. ip vrf 

The trigger of the BGP crash is a result of a racing condition between event 1 and event 2.

and the workaround is :

Workaround:

Since this is a racing condition, the workarounds are: 1. Not applying (1) before (2). 2. Give sufficient time for (1) to complete before applying (2). 

(1) no address-family ipv4 vrf  (2) no ip vrf  (3) ip vrf 

Known Fixed Releases:

12.2(33)SRE3 or later

However I am wondering how long is that "sufficient time" for (1) to complete? and how can I be sure if I can continue to next step(2)?

I collected some more info about the crash and I found that the last four Processes were :

1. BGP Router
2. BGP Generate Update
3. BGP Update table
4. BGP Update Topology peers

and then it detected a Crash -> created a Crashinfo file -> reloaded.

I think I can use 'show process cpu' command to verify if the BGP Processes are stable and I can continue to next step...

I want to listen to your opinion, do you have any Ideas? or do you know the exect time? or any other command to verify the safty?

thank you and happy new year!!

[Philip D'Ath · Accepted solution]

This bug has been fixed for quite some time.

https://tools.cisco.com/bugsearch/bug/CSCtg94250

Rather than wondering how long to wait in case it crashes wouldn't it be better to apply software that has the bug fixed?

The "gold star" release for platform looks like it is 15.3(3)S6.  I would change to running a gold star release.

https://software.cisco.com/download/release.html?mdfid=282201760&catid=268437899&softwareid=280805680&release=15.3.3S6&relind=AVAILABLE&rellifecycle=MD&reltype=latest

[loveandrew821]

hi, thank you for your kindness

I 100% agree with you that upgrading would resolve all the problem.

However, my boss wants to figure that out and we have some reason that we are not available to upgrade the IOS.

I was thinking about it, and I think it is not possible to know the exect time because of the differnce of the size of site and configuration. so, I have to know the command to verify the Processes or any other way to be sure that the Processes are stable.

[Philip D'Ath]

Two easy solutions comes to mind:

1. Don't delete VRFs any more.  Just leave them there.
2. If you must delete them, do it the next day.  It is not an urgent task.

[loveandrew821]

Can you please provide the reason why I should delete VRFs next day?

I don't believe there is any reference or docs about this Issue, but I'd like to listen to your opnion. Then let me investigate deeply based on your opinion.

[Philip D'Ath]

You were asking for the "exact time" you had to wait.  A safer option is to just defer the deletion till the next day.