Poor CDE response

[trisco]

90% of the time, CDE is giving poor response while: 1)starting 2)Selecting File Mgr. 3) Selecting Style Mgr. This is happening at both the console (lft0) and a remote terminal (dtterm). The terminal session response within CDE is good 100% of time. Since the console is direct attached, this would rule out any network issues, I think. No changes have been bade to the vanilla installation of AIX 4.3.3. I'm running on an RS/6000 - H80.

Any suggestions, would be appreciated.

 

[PSD]

trisco,

Are you using DNS?? or NFS??

Do you have any ML levels installed,? check via:

instfix -i |grep ML

Post this back to me.....

Cheers

PSD
HACMP Specialist

 

[trisco]

We're running DNS on another platform, but at this time I don't think that AIX is participating. Just using /etc/hosts. I thought that ML 06 was installed, but the instfix command displays only the base, 4.3.3.0 and no ML's.

 

[PSD]

trisco,

I have had a look at IBM webs site for APAR's and cannot see anything which relates to this problem - I would suggest going to latest ML i.e. 6 to see if the fixes make any difference. If you log a call with IBM that is always there first suggestion anyway.....

To see if you are using DNS does the /etc/resolv.conf file exist?

In terms of the problem if it was just slow when logging into CDE it could be a network/DNS/NFS issues i.e. due to lack of response, but as you have slow response from file and style manager this sounds particularly strange.

Cheers

PSD
HACMP Specialist

 

[trisco]

Thanks. I will upgrade to the latest ML that I have, which is level 6 and see if that helps.

The /etc/resolv/conf does exist and contains a definition to the name server.

Because of the poor response from File and Style manager, I too don't suspect a network issue. But I am keeping all bases as suspect.

 

[PSD]

trisco,

Can I just ask - does telnet take a long time to respond when telneting from the H80 to another server using DNS lookups???

Do you use a /etc/netsvc.conf file??

If so post it here for me to have a look at.

Thanks

PSD
HACMP Specialist

 

[trisco]

We have 3 h70's beside the h80. CDE and telnet has a slow response time on 2 of the 3 h70's. The one that has good response, both CDE and telnet, has a /etc/netsvc.conf containing:
hosts=local,bind4
The other 2 h70's and the h80 do not have a /etc/netsvc.conf file.

I think you are on to something, PSD, thanks. I will create a /etc/netsvc.conf with hosts=local,bind4 on one of the other h70's and will post the results.

 

[trisco]

Creating the /etc/netsvc.conf on each machine did not improve the processes (telnet & CDE).

I happened to remember that we had a similar issue with the one machine providing the better response. It is "Reverse Name Lookup". Per IBM on this earlier issue:
If bos.net.tcp.client level > 17
then add the -c option to /etc/inetd.conf for the telnet and ftp.
refresh the inetd daemon, by "refresh -s inetd" command.
This process bypasses the 'reverse name lookup', and improves the telnet and CDE processes.

If there is something else that I need to know, please post it here.

Thanks so much for your help.

 

[PSD]

trisco,

Do you have the client fileset above 17 on the slow machine??

I think that applying ML-06 is the first thing to do and then may be look at n/w card configuration.....i.e. entstat en0 and also lsattr -El ent0....

Cheers

PSD

 

[trisco]

My client fileset is at or above 17, and all filesets are at ML 06.

My CDE and telnet response has greatly improved since adding the -c option to FTP and telnet. But this still isn't clear to me how it affected the CDE response. Maybe there is something funky in my n/w card configuration.

Thanks PSD.
Trisco