System unreachable in policoes

Hi! When some nodes are unreachanble, it seems that NetBackup 5.1 (MP3A, Windows 2003) is attempting to connect to various ports on theses servers during about 45 minutes (per node unreachable). The consequence is that the policy in which these nodes are included can start several hours later than planned :-(

Is that the expected behavior?
Is there a way to specify a shorter timeout?

Cheers, Michel.


Extract from bpsched log. NB tries to connect to each node unreachable during 45 minutes it seems! Then, it submits the jobs for the backup of this policy:

...
09:10:29.474 <2> nb_getsockconnected: Connect to gvaclu04 on port 602
09:10:51.459 <2> nb_getsockconnected: Connect to gvaclu04 on port 707
09:11:13.443 <2> nb_getsockconnected: Connect to gvaclu04 on port 702
09:11:35.427 <2> nb_getsockconnected: Connect to gvaclu04 on port 696
09:11:57.411 <2> nb_getsockconnected: Connect to gvaclu04 on port 727
09:12:19.505 <2> nb_getsockconnected: Connect to gvaclu04 on port 738
09:12:41.489 <2> nb_getsockconnected: Connect to gvaclu04 on port 540
09:13:03.473 <2> nb_getsockconnected: Connect to gvaclu04 on port 522
09:13:25.458 <2> nb_getsockconnected: Connect to gvaclu04 on port 859
09:13:47.442 <2> nb_getsockconnected: Connect to gvaclu04 on port 654
09:14:09.536 <2> nb_getsockconnected: Connect to gvaclu04 on port 880
09:14:31.520 <2> nb_getsockconnected: Connect to gvaclu04 on port 645
09:14:53.613 <2> nb_getsockconnected: Connect to gvaclu04 on port 552
09:15:15.707 <2> nb_getsockconnected: Connect to gvaclu04 on port 657
...
2 Replies

Re: System unreachable in policoes

HI,
In the NB configuration we give a range of client ports
the default is 512 to 1023

While performing client backups NB uses a port from these range.
Also the selection is random if the ":use random port selection is enabled"

is there a way to specify a shorter timeout?

- What are the timeout settings in the Master server properties.

Re: System unreachable in policoes

Well... that's an idea, but before restricting the client port range to something smaller, I need to better understand the impact! Currently, I have the default 512 to 1023. It seems to try ~120 random ports and/or a timeout of 45 minutes; then it gives up for this client!

Tiemouts on the master servers are default, except for Client Connect and Client Read timeout (set at several hours. Needed when you have large SQL DB to restore).