Threads going to Error Status

[Marsha Parks]

I have noticed since we upgraded to 5.8.4 that several of our interfaces will go to error status when the remote server temporially goes away.  Under the old version, we would stay in opening and wait for the server to come back.

Has anyone else noticed this change and have any suggestions on how best to change my configuration that I can stay in opening and not filp to error status?

Marsha Parks

Topic

[Russ Ross]

I’m curious if the problem seems random or specific to certian interface(s).

If cycling the interface on cloverleaf causes the thread to go back to opening then an alert to recycle the thread on ERROR or not UP could be a workaround while searching for the underlying cause.

Almost all of our interface in prodcution have an alert to notify (email/page) and cycle an interface when not UP.

I have not seen the error issue you describe that I’m aware of but have seen interfaces think they are connected and UP when they aren’t.

Our alerts for outbound interfaces to check for number of resends has taken care of this nicely and our alerts for last received helps some with inbound interfaces.

Russ Ross
RussRoss318@gmail.com

[Bob Richardson]

Greetings,

We are evaluating 5.8.5 on AIX Unix (VM) TL 7 and have noticed that our few Cloverleaf configured FTP connections can go to ERROR status when the remote is unavailable.   Usually we set our (and we only have a few of these) FTP threads to “delay connection until needed”.  The only other instance of going to ERROR would be if you configure a TCP/IP thread to connect only when needed using Driver Control metadata key to wait for a reply and the remote disconnects (we did NOT implement any of these but experimented with it and requested an enhancement to just put this type of thread to Opening status).

We will be interested in your findings on this issue in your shop.

Thanks for contributing!

[Marsha Parks]

It has been specific interfaces and I have been able to link it to one of our teams rebooting a DNS server.  So while it is a short time for the reboot, the interface goes to error status.  

I am good with using our alerts to monitor and a script to restart.  I was just wondering if anyone else noticed this change from 5.5 platform that the interface would stay in opening status and now in 5.8 we go to error.  

Marsha Parks

[Russ Ross]

One more reason to put a smile on my face about using IPs instead of DNS machine names.

Our DNS servers have hurt me bad in the past and I said no more and it sure has been much better since using IPs instead of host names.

The comments about FTP reminded me about a gotcha I was able to fix.

We have some filesystems exported and NFS mounted between boxes.

I have observed if I do an HACMP fail-over on our AIX cluster I need to recreate the NSF export list to have NFS mounted files systems to be useable on the other foreighn systems.

Here is a code snipet from my HA script I wrote to illustrate the command that takes care of this on startup after a fail-over

Code:

#————————————————————————-
# export all the desired directories defined in /etc/exports
# which will end up being deifined in /etc/xtab after running this command
#————————————————————————-

/usr/sbin/exportfs -a

Russ Ross
RussRoss318@gmail.com

[Author]

Replies