Recoverying from a dead gateway?

[Doug O'Leary]

Hey;

I have a client who had dead gateway detection enabled and had a network glitch which kicked off the whole "I'm not using that route anymore, ever!" thing.  Short of a reboot, how can I recover from that?  I tried cycling the lan card and deleting/re-adding the route but neither of those worked.  Since the client was overy my shoulder, tapping his foot, I finally rebooted so the box is up.  It just seems like that's killing a fly with a small thermonuclear device... 

Thanks.

Doug O'Leary

P.S. This thread has been moved from HP-UX > System Administration to HP-UX > networking - Hp Forum Moderator

[Patrick Wallek]

>>I tried cycling the lan card

What did you do here?  'ifconfig down' & 'ifconfig up'?

Perhaps a '/sbin/init.d/net stop' and /sbin/init.d/net start' would help.

[Doug O'Leary]

Hey

Thanks for the reply. Yea, it was ifconfig lan0 down; ifconfig lan0 up. I also tried the /sbin/init.d/net stop&start w/o success as well.

You wouldn't think this would be so hard...

Thanks again.

Doug O'Leary

[Patrick Wallek]

If you have access to the patch database (ie. a software support contract linked to your ID) have a look at the information in the WARNINGS section of PHNE_42470.  It is for 11.31 but the same thing may work on older systems.

[Bill Hassell]

>> dead gateway detection...

Turn it off on every HP-UX system you have. I have yet to hear from anyone that thinks this default setting is useful. It will trash SG clusters and cause desparate sysadmins to TOC the computer (with nasty side effects) because the system is 'hung'.

[Aneesh Mohan]

I would like to share this with you , there is a known issue with PHNE 42470 ARPA patch , the work around in mentioned in the bottom.

A Patch Warning has been issued for PHNE_42470 :

11.31 cumulative ARPA Transport patch

You are receiving this notification because you either downloaded the above named patch or it was downloaded on your behalf. The patch was sent to the following company:

Job-specific information for Job 44457):

• Company: National Commercial Bank
• Customer Name: Adnan A. Al-Herz
• Customer Phone: Not specified
• Customer Email: xxxxxxxxxx
  
• SAID: xxxxxxxxxx (For the security, Email address and SAID were removed - HP Forum Moderator)

The following information has been added to the patch documentation for PHNE_42470 :

13/05/27 - This Non-Critical Warning has been issued by HP.

- PHNE_42470 introduced behavior where the route(1M) command

may fail with the following message:

Network is unreachable

- This behavior only manifests itself if an entry

already exists in the route cache table.  The behavior

is observed regardless of the type of route, host, net

or default being added.

- The following approach can be taken to work around

this behavior:

- bring the related interface down using:

ifconfig lan<x> down

- run the command:

ndd -set /dev/ip ip_ire_gw_probe 0

- bring the related interface back up using:

ifconfig lan<x> up

- add the desired route

- run the command

ndd -set /dev/ip ip_ire_gw_probe 1

CAUTION: Observe that network connectivity over the

involved interface will be lost between the

"ifconfig ...  down" and "ifconfig ...  up"

commands.

- Additional details on this behavior may be found in Change

Request QXCR1001280056.

- This behavior will be corrected in the next cumulative

ARPA Transport patch to be released as soon as possible.

Regards,

Aneesh

[Doug O'Leary]

Hey, Bill;

Thanks for the reply. I've been trying to come up with a scenario in which this setting would be a good idea and the only thing I could think of is if the host has multiple valid routes. I tend more towards the theory that "routers route and servers serve"; but, maybe this is a requirement for that ever present edge case. I agree, though, it very definitely shouldn't be the default and I do turn it off on every system I manage.

Thanks again.

Doug

[Doug O'Leary]

Hey, Aneesh

Thanks for the reply. So, if I'm reading that correctly, the work around is to:

1. delete the route
2. disable the lan card
3. turn off dead gateway detection
4. enable the lan card
5. re-add the route
6. turn on dead gateway detection (if you're silly enough to want to do so...)

If that's actually the right answer, it sounds like a reboot is easier. Yay for the microsoft school of troubleshooting.

Thanks again, all. I appreciate the help.

Doug O'Leary

[Bill Hassell]

>> host has multiple valid routes...

But if the router stops respondingf to a ping, it is either defective or the ping is turned off. In either case, disabling the route means that the system will not recover when the router is once again functional. If a route is defined, the system should continue to use it or retry when the route doesn't work. The setting might possibly be used to stop logging "no route to host" messages, but if should not be the default. 

My $.02 USD worth...

[turboexcesx]

1. delete the route
2. disable the lan card
3. turn off dead gateway detection
4. enable the lan card
5. re-add the route
6. turn on dead gateway detection (if you're silly enough to want to do so...)

it Worked?

can you translate it to commands?

Thanks you!