Lost SSH connections

[Cretin]

we are running Oracle Linux 5.2 on platform x86_64. We are running Linux natively on the hardware. We have 2 Oracle grid servers. Grid1 is the production server and gridDR is the disaster recovery server. Grid1 constantly sends archive logs to GridDr. At some point GridDR lost SSH connectivity and could no longer recieve the archive logs. That caused the space on Grid1 to fill up and Grid1 became unresponsive. The Linux administrator says there is nothing in the logs to indicate why GridDR may have lost SSH the logs only state that SSH connectivity was lost. We are using OEM 10G for grid. Any ideas what to look for or where to look to try to determine why GridDR lost SSH connections?

Cretin

 

[Cretin]

More Information, we could ping the server but yet it was accepting no connections nor was it able to establish a connection. Both the Linux administrator and the DBA manager stated they have never seen that before.

Cretin

 

[Annihilannic]

How was it resolved? Reboot? Restart of SSH daemon?

I think it'll be hard for us to guess what happened without being able to see the system in that state really... but the state of the SSH daemon is the first thing I'd have checked, i.e. whether it still accepted connections locally (which would indicate a network issue), or not at all. If it was fine I'd have started it in debug mode to get some more info. I'd have also tried running the ssh client in verbose mode from the other server to see which stage the connection was failing at... although this is rarely informative.

Annihilannic
[small]tgmlify - code syntax highlighting for your tek-tips posts[/small]

 

[stefanhei]

Do you move the archivelogs "manually" (i.e. via script)? Does gridDR host a standby database? If my memory does not fail me the standard way of logshipping does not use ssh but the Oracle Net protocol. SSH is only used for the deployment of the Management Agents.
Anyway, do you see lots of CLOSE_WAIT or FIN_WAIT (or any *_WAIT) when you issue

netstat -an

?