Re: Node in a 2 cluster RAC environment keeps failing over

  • From: Mogens Nørgaard <mln@xxxxxxxxxxxx>
  • To: oracle-l@xxxxxxxxxxxxx
  • Date: Fri, 19 Mar 2004 19:58:59 +0100

In the bank I worked in from 87 to 90, Peter Gram and the other super-nerds would take in the Olivetti M24 PC's where the hard drive would constantly freeze and caaaaarefully lift up the right hand side of the unit about 10 cm's off the table, then let it drop. That would always free the drive.

What was funny was when end users called in with the problem. The guys - the Peter Grams of the world - would then instruct them over the phone:

PGs: "Please lift - VERY CAREFULLY - the unit in the right hand side about 10 cm's up in the air. Careful!"
User (excited, scared): "OK, I'm HOLDING it now..."
PG's: "Now release it - drop it!"
User: "What?!?!?"
PG's: "Drop it. Let it go. Suddenly!"
<WHUMP>
User: "It's done. What now?"
PG's: "Try to start up the PC again..."
User: "It works!"


Peter Gram knows many such stories.

I bought an M24 SP (Special Performance - 10 MHz instead of 8, 20 MB harddisk, 640 MB RAM, monochrome) in 1986 just before joining the bank. It cost me - believe it or not - 42K Danish Kroner. Of course. I think I finished paying for it a couple of years ago.

Mogens

Paul Drake wrote:

--- Mladen Gogala <mladen@xxxxxxxxxxxxxxx> wrote:

Ask them to perform the "drop test". Here's the URL
describing the test:



http://www.computerworld.com/departments/opinions/sharktank/0,4885,91405,00.html



Hey - when I worked at a ComputerLand store in the early 80s, the recommended height to "fix" an Atari via a gravity check was 36 inches, none of this one foot stuff.

Of course, you had to take the unit back into the
repair room and close the door, lest the customer hear
the sound of the impact ;)

Boom. That will be $30 please.

Pd


On 03/19/2004 09:48:30 AM, laura pena wrote:

Thanks Lee,

So frustrating. Veritas has been working on this for the past 3 to

4


months now. They currently say the VxConfig Daemon

is


crashing and have escalated the bug.

But before they said gabconfig was issue.

Thanks for the information.

--- Lee Jenkins
<lee.jenkins_at_remotedba.co.za@xxxxxxxxxxxxxxx>
wrote:

Hi Laura,

I think the set-nofastpath in my ltttab is
specifically for a bug in HP-UX.


You should probably pass it by Veritas support,

but


I would set the network cards to auto-negiotiate off, and set to both
100Mhz. (2G is overkill)


Do both node have the same 100 MHz/1G setup and

are


plumbed together?i.e. 100->100MHz and 2G->2G?

Our client setup is 2 private heartbeats

(100Mhz)


and 2 public LANs (1G). This way either network card can fail, and
everything still runs.



So if I understand we should try setting

set-nofastpath 1 in our /etc/llttab so it will

use


the slower connection to perform a heartbeat

check?


Yes.

Regards,
Lee


laura pena writes:


Wow someone who has heard of this.

Yes, we have our interconnects set to etherfp.
LLT link information:
Link Tag State Type Pri SAP



MTU Addrlen

Xmit Recv Err



LateHB

Broadcast
0 eri0 on etherfp hipri

0xCAFE 1500 6


261840450 240300348 0



89

FF:FF:FF:FF:FF:FF
1 ce1 on etherfp hipri

0xCAFE 1500 6


261914713 240332338 0



91

FF:FF:FF:FF:FF:FF

we have 2 interconnects but they are of

differnts


speeds:

set-node VLDBN1
set-cluster 10
link eri0 /dev/eri:0 - ether - -
link ce1 /dev/ce:1 - ether - -

I believe ce1 is the 2 gigabit and eri0 is the 100M


This setting was done by our SA. We do have a

ce0


which is another 2 gigabit, but our SA wants to
leave this for public connection. Do have a

similar


setup?



So if I understand we should try setting

set-nofastpath 1 in our /etc/llttab so it will

use


the slower connection to perform a heartbeat

check?




Thanks so much for responding.


-Lizz






Lee Jenkins <lee.jenkins@xxxxxxxxxxxxxxx>

wrote:


Hi Lizz,

I posted this response on oracle-l, but it

doesn't


seem to have appeared.

On a customer of ours 3 node cluster VCS v3.5

(HP-UX), we had to set "set-nofastpath 1" in /etc/llttab so as to force the heartbeat to

operate


at 100Mhz. Have you got 2 heart beat

interconnects?


i.e. redundancy?

You can get stats by running lltstat, which

shows


data volumes and errors.

You can run lltstat -l to check the setting of

fastpath. "ether" means fastpath is disabled,
"etherfp" means enabled.


Regards, Lee


Lee Jenkins
RemoteDBA
www.remotedba.co.za
Tel: 011 447 0533
Fax: 011 447 0533
Cell: 083 408 0857



Do you Yahoo!? Yahoo! Mail - More reliable, more storage,

less


spam



__________________________________
Do you Yahoo!?
Yahoo! Mail - More reliable, more storage, less

spam


http://mail.yahoo.com


----------------------------------------------------------------

Please see the official ORACLE-L FAQ:

http://www.orafaq.com


----------------------------------------------------------------

To unsubscribe send email to:

oracle-l-request@xxxxxxxxxxxxx


put 'unsubscribe' in the subject line.
--
Archives are at

//www.freelists.org/archives/oracle-l/


FAQ is at

//www.freelists.org/help/fom-serve/cache/1.html


-----------------------------------------------------------------

----------------------------------------------------------------

Please see the official ORACLE-L FAQ:
http://www.orafaq.com


----------------------------------------------------------------


To unsubscribe send email to: oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at
//www.freelists.org/archives/oracle-l/
FAQ is at
//www.freelists.org/help/fom-serve/cache/1.html



-----------------------------------------------------------------



__________________________________ Do you Yahoo!? Yahoo! Mail - More reliable, more storage, less spam http://mail.yahoo.com ---------------------------------------------------------------- Please see the official ORACLE-L FAQ: http://www.orafaq.com ---------------------------------------------------------------- To unsubscribe send email to: oracle-l-request@xxxxxxxxxxxxx put 'unsubscribe' in the subject line. -- Archives are at //www.freelists.org/archives/oracle-l/ FAQ is at //www.freelists.org/help/fom-serve/cache/1.html -----------------------------------------------------------------


---------------------------------------------------------------- Please see the official ORACLE-L FAQ: http://www.orafaq.com ---------------------------------------------------------------- To unsubscribe send email to: oracle-l-request@xxxxxxxxxxxxx put 'unsubscribe' in the subject line. -- Archives are at //www.freelists.org/archives/oracle-l/ FAQ is at //www.freelists.org/help/fom-serve/cache/1.html -----------------------------------------------------------------

Other related posts: