EMS detect a failure and don't TOC the package

[Luciano Avelino_2]

Hi,

Is there any configuration on resource_manager using EMS with MCSG to detect a PV/LVM failure and just notify the system administrator without request A TOC of package ?

tks in advance

Luciano

[Juan M Leon · Accepted solution]

If the system is TOC means that none of the services are able to run, on any node. What errors are you getting on your cluster log files.
I dont believe that EMS will TOC the server,
to check status you can execute the following
resls /vg/vglock/lv/status

Also I will personaly think that a SECOND_CLUSTER_LOCK is always helpfull

I hope this helps

[Juan M Leon]

I Think I read your question incorrectly, there is a option with in your cluster configurations to have the node TOC if problem was found.

[Luciano Avelino_2]

sorry, I used the wrong word ( TOC ) , I need know if there any way to configure a resource manager in may package of MCSG to just send a notification and not at all.

tks

Luciano

[Pedro Cirne]

Hi Luciano,

I don't think a disk failure will cause SG to TOC. On a standard configuration it won't!

But, a disk failure will be detected by EMS and root will receive a mail with that information. You may also configure EMS to send a message to OVO if you have it.

Enjoy :)

Pedro

[Luciano Avelino_2]

Pedro, tks for your answer

but the package of MCSG will be the resource manager with resource dependence, so when this dependence ( resource manager ) detect a hardware failure, it will ( resource manager ) go down , and the package will failover. it's ok ?

I need the package not failover , just send a notification e-mail, it's possible ?

tks in advance

[Jeff Schussele]

Hi Luciano,

MC/SG will not act on disk "failures" because how would it know whether it's actually "down" or just horribly overloaded?
And if it is - in fact - dead, MC/SG still cannot deal with it because it needs to sync it & dismount it which it obviously couldn't do if it's dead.
The best you could do with MC/SG is to set up application monitoring such that if the app is getting no respone on disk I/O then perform some action - like e-mail someone or stop the application. And the latter may be problematic if there is no disk I/O.

Only OnLine Diags (EMS) can & should do this.

My $0.02,
Jeff

[Luciano Avelino_2]

Jeff,

The MCSG don't know when disks is in failure condition ( I know ), but the EMS know , and the MCSG take a failover when it happens .. and don't want this condition , I need just an e-mail with notification.

Luciano

[Luciano Avelino_2]

Mrs,

I have two SD64000 in a high availability enviroment with MCSG. My oracle database is in a package , and I want put a resource manager using EMS MIBS to monitor the MirrorUx health on VGs of my package, I know when one of two disks in may mirror get failed , the resource manager will get down , and may package will failover , I need know if there any thing that I do to EMS send just an e-mail massage and not send a failover instruction to package.

tks in advance

Luciano Avelino

[Jeff Schussele]

Ok - then you need to know how to use monconfig.
man it for details.
Odds are it's already OK - but check it none-the-less.
Also you'll need to set up the /etc/mail/aliases file so that root mail will be properly routed to you or whomever.
Now as to why the packages TOCs (and I'm not sure what you mean by that), we'd need more detail such as 1) What disk? (vg00, application, etc) 2) Is the package switching or the system rebooting?
If it's a vg00 disk then the SYSTEM is TOC'ing not the package & you'll have no control over that. If it's an application disk them it may be MC/SG monitoring & in that case then just edit the script such that it won't "TOC" in those circumstances.

Rgds,
Jeff

[Luciano Avelino_2]

Jeff,

I know how to use the monconfig and know how to configure the e-mail aliases to send external messages and I know that I cannot put the vg00 in a package. The question is :

When the resource manager in my package get down ( a disk failure , a mirror failure , a PV link failover , or any other causes ... this is not important ) , the package will failover to another node ( the database will shutdown on node A and start on node B ) , there is any thing that I can do to just send an e-mail message and get the package intact ( without move the database from node A to B )

tks

Luciano

[Jeff Schussele]

OK - as it appears you're using the EMS/MCSG feature then I'll boil it down to two words:
package dependency.
You need to remove that/those disk/s from the dependency - MC/SG it just doing what it's told. Or you need to fisk the disk "problem".

Ciao,
Jeff

[Luciano Avelino_2]

Jeff,

I will think about , maybe I will put the EMS running without package.

tks a lot

Luciano Avelino

[Luciano Avelino_2]

tks a lot