Troubleshooting
Problem
When restoring specific files from a large backupset, the server needs to read all backupset tapes. The client session may close with ANS1017E before the needed file(s) could be restored.
Symptom
The following is the error as reported in Tivoli Storage Manager Command Line:
[root@hostxxx]:/ dsmc restore backupset /fs/path/filexx /fs/restorepath/ -backupsetname=TEST-BKSET.44444444
.....
Restore function invoked.
ANS1247I Waiting for files from the server...
After some time ...
>>>>>> Restore Processing Interrupted!! <<<<<<
ANS1017E Session rejected: TCP/IP connection failure
[root@hostxxx]
On Tivoli Storage Manager server side, no error or timeout messages are reported.
Cause
A system or firewall timeout closed the session before it could complete
Environment
Client restoring specific files from a large backupset without using TOC.
Diagnosing The Problem
Check if the client session is not closed due to any Tivoli Storage Manager server timeouts. In the server activity log, search for message like:
ANR0481W Session session number for node node name (client platform) terminated - client did not respond within commtimeout seconds seconds.
or
ANR0482W Session session number for node node name (client platform) terminated - idle for more than idletimeout minutes minutes.
If those messages do not appear before the time of the ANS1017E on client side, then the server is not responsible for the session closure, so this technote may apply.
Resolving The Problem
The solution is the same than the one documented in the technote shown in the URL section. In the other technote the problem is a hang of the client session while here a session lost error is reported instead of a hang situation
Tivoli Storage Manager uses the SO_KEEPALIVE network socket option. The SO_KEEPALIVE option allows Tivoli Storage Manager to enable periodic transmissions on a connected socket, when no other data is being exchanged.
The solution is to decrease the SO_KEEPALIVE interval to allow the session to be kept alive even if no transmission happen.
The default keepalive interval can be updated to a different value. For example, to set a keepalive interval of 10 minutes, the following command can be issued :
- on AIX, "no -o tcp_keepidle=1200"
- on Solaris, "ndd -set /dev/tcp tcp_keepalive_interval 600000"
- on Linux Redhat "sysctl -w net.ipv4.tcp_keepalive_time=600"
- on Windows, use Registry Editor (at your own risk) to modify/create value data for the "KeepAliveTime" DWORD value to 600000 in the following registry key: HKLM\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters
Related Information
Was this topic helpful?
Document Information
Modified date:
17 June 2018
UID
swg21957993