Re: Help required for Tru64 Unix Alpha

[Sameer Patel]

Hello,

1. Today, we could not ping to our COMPAQ ALPHA SERVER (TRU64 UNIX 5.1B with ORACLE 9.2.0.4) and the message when seen in mail was:
EVM daemon: High event activity exceeds $count in $period minutes.
We had changed earlier the parameter event from 500 to 750 in evmdaemon.conf. There are no messages in /var/adm/messages. The ORACLE was also running fine.
2. Also, in the evmdaemon.log files, we found the following:
socket binding failed, errno-49.
could not open remote accept socket.
3. The evmshow command gave following:
smsd: starting to log...
failed to vol_check().
successfully bound to port 596.
ready for client connections.
cannot set parent conenction. object does not exist.
cannot register service:rpc:unable to send; errno=no buffer space available.

Request your help.

Thanks & Regards,
Tejas.

[Ralf Puchner]

1. check the event information itself (evmshow) be sure what kind of event will be repeated so often. See also http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=488671

2. seems a socket can not be initialized or all sockets are used (netstat command)

3. vol_checks indicates a vold problem (lsm).
Please check your setup

[Sameer Patel]

Dear Ralf,

Can you kindly elaborate on points 2 & 3. I have also increased threshold from 750 to 1000 in /etc/evmdaemon.conf. Will that solve my problem of point no. 1?

Regards,
Trupti.

[Ralf Puchner]

No why should this solve a problem? You only change the event-highwatermark intervall nothing else!

Please check what kind of event leads to the messages...

Regarding 2&3: use the given command and check lsm configuration as described within the lsm troubleshooting section of the manual. Maybe problem 3 is a follow on problem of 2.

[Sameer Patel]

Hello,

Today, suddenly at 3:45 pm, we could not ping or telnet to the server. Oracle on the server was up. Infact, login from the console was also disallowed. It gave the following message:

The DT messaging system could not be started.
To correct the problem:
1. Choos ok to retunr to the login screen.
2. Select Failsafe Session from the login screen's option menu & login.
3. Check to see that your hostname exists correctly in /etc/hosts if your network has already been configured.
4. If your network has not yet been configured, make sure that /etc/hosts has the following entry in it:
127.0.0.1 localhost.
For additional information, see the CDE user's guide.

When I check through ifconfig -a, I get both the ee0 and ee1 interfaces up. When I ping localhost, I get the following:
ping:sendto: No buffer space available.
ping:wrote:192.168.2.203 chars, ret=-1.

I am attaching herewith the /etc/hosts file and also the events occured.

Request your help URGENTLY.

Thanks & Regards,
Trupti.

[Ralf Puchner]

maybe time to think about professional service. a cde problem is typically if network will not work, so don't care about it.

As I can see from the given event output there are lot of errors and cluster seems to be setup a little bit "funny".

Machine reported advfs errors, too much open files and maybe socket problems. Not a good starter for a machine, so be sure there is no port scanning or defective hardware in the network. Btw. a netstat -i will be a starter to analyze such problems, but as I've written before please open a call within the HP support center to get professional help for all of your problems!