Operating System - HP-UX
1748052 Members
4814 Online
108758 Solutions
New Discussion юеВ

Re: Dispatchers going down in Oracle 7.3.3

 
Timothy Czarnik
Esteemed Contributor

Dispatchers going down in Oracle 7.3.3

Hey all,

Its crisis time here. Here's the scoop. For the last 3 weeks now, on a daily basis, the dispatchers in Oracle have been failing. At that point, the listener logs that a connection is coming in, but thats where the whole process stops.

Anyone seen this before?

Anyone have any ideas?

Before we start, I know this version of Oracle is no longer supported. I also know that we should be on a newer version. Thats coming, but it isn't a reality yet.

HP-UX 10.20. Patched to June 1999 ( I know! I know! Its an inherited box!). The thing was, up until 3 weeks ago it was running smoothly. Looking for a miracle here, and will reward points like a French Skating judge to get one! :)

-Tim
Hey! Who turned out the lights!
7 REPLIES 7
Volker Borowski
Honored Contributor

Re: Dispatchers going down in Oracle 7.3.3

Hello,

I'd like to see the init.ora parameter concerning the MTS configuration.
Did you install new clients 3 weeks ago (where the Oracle client SW might not match your server version) ?
It the drop reproduceable and related to the client issuing the connect (Check the IP in the listener log) ?
Are you on the edge of sessions when the drop occurs ?
Is the entrie database down when the drop occurs or are you still able to connect dedicated or BEQ (internal) to query some data ?

Just a start.
But if it started 3 weeks ago.... what changed since then ?

Good hunting
Volker
Timothy Czarnik
Esteemed Contributor

Re: Dispatchers going down in Oracle 7.3.3

Volker,

Thanks for the response. The weird thing is, nothing has changed. The night it started happening, we lost a network switch that the server was plugged into. When the switch went down, we swapped network cables with a known good cable until the switch was replaced, and then we went back to the old cable again. Since then, we've done a number of shutdowns, one of which was a full shutdown/power off in the hopes of resetting something.

This is weird. I'm hoping that someone has seen this before.

-Tim
Hey! Who turned out the lights!
A. Clay Stephenson
Acclaimed Contributor

Re: Dispatchers going down in Oracle 7.3.3

Hi Tim:

In that case, I would make sure that the network settings on the switch are hard set to 100FD (or whatever) and make sure that the settings on your host(s) are also hard-set. Remarkably, mis-matched setting will work just well enough to drive you nuts.

Regards, Clay
If it ain't broke, I can fix that.
Timothy Czarnik
Esteemed Contributor

Re: Dispatchers going down in Oracle 7.3.3

Clay,

Excellent idea. Its been verified. Both are hard set to 100FD.

Thanks!

-Tim
Hey! Who turned out the lights!
Volker Borowski
Honored Contributor

Re: Dispatchers going down in Oracle 7.3.3

Bingo

if that switch changed, this is the reason.
Get tight to it.

Check syslog on the switch !
Read the fault and collision tables.
Is this switch cascaded to another one ?
How is it uplinked ? 100 MBit FDX Hardcoded ?

Keep in mind, that some switches are very sensitive against change of mac-adresses which deliver the same IPs (which might be the case with your replaced switch).

Is the replaced switch from the same vendor / same modell ?

Volker
Timothy Czarnik
Esteemed Contributor

Re: Dispatchers going down in Oracle 7.3.3

Volker,

This is what we thought as well, but of course the network side of the house doesn't agree 100%. They put a sniffer on the network, and can see packets going into and coming out of this server. When checking for the individual ports that the dispatchers are using, things run along fine, then all the sudden packets go in but never come out.

They say that this proves that its not a network issue, because the packet gets to us. I can't say as I disagree with them about that, so I will continue looking into this.

-Tim
Hey! Who turned out the lights!
Volker Borowski
Honored Contributor

Re: Dispatchers going down in Oracle 7.3.3

Tim,

if

... "When the switch went down, we swapped network cables with a known good cable until the switch was replaced" ...

this cable did the job well (on another switch obviously), I would switch back to this configuration as a debug-starter (whatever any sniffer would say).

Then use exactly this cable to the port on the new (replaced) switch.

Did you look into Oracle alertfile and search for written tracefiles in usertrace ?

Volker

(I'am no skater, so 2 times 7 is fine for me !Until we get real results 3 will do fine :-)