request socket connection not authorized: 15

[Enrico Venturi]

Hello colleagues,
what can be the reasons of the error shown in the title?
the available sockets are exausted?

thanks
Enrico

[Jerome Baron]

Hi Enrico,
can you explain the context which give you this error ?

Regards,
Jerome

[Enrico Venturi]

Ok, in my platform there's HP OpenView NNM; I run a lot of processes so that about 750 sockets are engaged; at a certain moment the ovspmd (daemon of OV DM platform) stops accepting any more requests: the error is
ovstatus: No response from ovspmd. Check ovspmd.authfile on server xxxx.
By taking a look on nettl trace file I see the mentioned error.
It seems that the ovspmd doesn't accept anymore incoming requests. As soon as ai stop some process (and it frees its sockets) the ovspmd restarts accepting requests.

[Ian_4]

(a) See http://openview.hp.com/sso/ecare/getsupportdoc?docid=S3100005790

In this particular case 'ovspmd.auth' has a short name and when the
fully qualified name is looked up it returns with the short name only.

A suggested resolution to this problem is to change 'nsswitch.conf' to
look at files first. Make sure the lookup works for ip, long name, and
short name. Then, kill the 'ovspmd' -newconfig process that is running
and run ovstart again.

(b) If the system is a multi-homed system then enables the NNM internal
process communication over loopback.

Here are the steps:
- ov.conf uncomment and enable
USE_LOOPBACK=ON

NNM_INTERFACE=10.10.10.10
IGNORE_NNM_IF_FOR_SNMP=ON

- ovspmd.auth uncomment
loopback +
localhost +
loghost +

- /etc/hosts, add the loopback entry:
127.0.0.1 localhost loghost loopback

- remove $OV_DB/openview/ovwdb/ovserver
- remove $OV_SOCKETS/OVsPMD_MGMT

Then run:

ovstart -d

Hope this helps.