10.2 running very slow

[Phillip Popp]

Hi All,
I have a weird problem I can not quite find the cause of. I am running 10.2 on a 745 box.
Every thing runs extremly slow. The hard drive was up around 97% and I brought it down to 87%, but it made no difference. I went into sam and looked at the processes taking up CPU time and it was 99% idle. Is there any way to check the memory, to see if there are any physical issues with it? I was told that this box was functioning at one time and was just not used for a while, but did function last time it was used.
Every thing seems to function correctly, it is just extremly slow. Like going into Sam: I log in and it takes a good two minutes or more for the main screen to come up. Only other thing wrong that I am told, is when it is rebooted, it does not go into the HP VUE mode. Not sure exactly what it does, do not believe it fails any of the system check's. I am running it remotle through telnet (which again is outrageously slow logging into). I had the person onsite do a control break to go into single user mode, from there I telented in, however the person on site is reporting same extremly slow responses as well.

Not really sure what to look for in the syslogs, I am still pretty new to the unix admin stuff.

Thanks,

Phil

[Pete Randall]

Phil,

I have to wonder if you're swapping. Run "vmstat 1 5" and take a look at the po column. If you see double digit numbers, you've got swapping issues and it will definitely mean slow going.


Pete

[Steven E. Protter]

Shalom Phil,

You need real data to look at.

http://www.hpux.ws/?p=6

That being said, the most common cause of Oracle performance problem is the enormous memory consumption of their software.

Areas to look at:
1) Kernel tuning. Sometimes exceeding minimums on filehands maxuprc and shared memory helps performance.
2) Adding physical memory.
3) Tuning the Oracle SGA to be less of a littl oinker (pig) on memory.

SEP

[Tim Nelson]

As mentioned start posting some outputs for stats.

sar 5 10
sar -d 5 10
vmstat 5 5
swapinfo -tam

We will start from there.

[Phillip Popp]

I am remote to the system and the onsite guy is in California for the next three weeks. I cant even log in currently, so I cant do anything with this yet.

Thanks for all the suggestions.

Phil

[Bill Hassell · Accepted solution]

The usual answer for very slow logins and failure of VUE (or CDE) points to networking, specifically DNS. A common culprit is a failure of the DNS server or worse, changes to the server that causes all sorts of machines to fail. The first thing to do is to change the file /etc/nsswitch.conf to use the hosts file first, then DNS. That will usually solve the slow login and possibly get VUE working again. The /etc/hosts file should have all the related machines listed and the change in nsswitch.conf is:

hosts: files [NOTFOUND=continue UNAVAIL=continue] dns

Another possibility is uptime shows massively large workloads, 10, 50 even 100. Since this is a single processor system, these large workloads are indicative of a problem with large page management. There is an adb hack to turn off this feature if the work load for uptime is way too high.

[Phillip Popp]

Thanks Bill,
Would the networking issue also cause slow operation once some one is logged in and just running routine things like sam?

I will definately look into that when I have a chance.

Thanks,

Phil

[Phillip Popp]

Thanks Bill,
Would networking issues also cause the system to run slow once you are logged in and doing routine stuff like running sam?

I will definately give it a look at when I get the chance.

Thanks,

Phil

[Pete Randall]

The main culprit when networking issues arise is a duplex mismatch between switch port and NIC. Use lanscan to find your NIC, then lanadmin to determine what speed/duplex setting it's using. Then find out what the switch is set to.

Of course, with a box that old, it might only be capable of 10MbHD.


Pete

[Amond]

Hi all,
I suggest too a "netstat -i" on the 745 to verify if you don't have collisions on your card. I had identified a deficient network component (bridge between two networks) by this way. I used then an netmon "digger" with HPOV to discover the pbs.

If there, try a local connection with telnet with a pc on the same lan.
If it 's works, it 's not your HP station, it's a network problem.
If it not works ,look at as says Phil, your network configuration on the HP computer.

Regards.
J.M.B.)

[Bill Hassell]

No, the networking (DNS) won't affect normal activity after yopu are logged in (things like vgdisplay or man terminfo, etc). Also make sure you have disabled *all* NFS in /etc/rc.config.d (3 files). NFS problems can kill your system performance. You can be sure that everything is working normally internally by using a *real* console - a 700/92 terminal or a laptop with a serial port (don't use the LAN console port unless the LAN cable is connected to an isolated subnet or better yet, peer-to-peer with a laptop).

What does uptime show?

[Phillip Popp]

Thanks Guys,
Bill had it right. I changed the nsswitch to his suggestion and it boots into HPVUE mode and all is well

Thanks again!!

Phil