LUN Device not showing without reboot on Redhat
 

Hi

I have a EMC CX500 Clarion connected to a RHEL ES 2.1

I have setup a new LUN but need to add this to the server without rebooting

I have read that you do the following

echo "scsi-qlascan" >>/proc/scsi/qla2300/<adapter-id>
echo "scsi add-single-device 3 0 0 3" >>/proc/scsi/scsi

But my adaptor ID is 'HbaApiNode' and was expecting it to be qlaXXX ?

Any ideas if I am on the right track here ..

Thanks

Mick

RHEL 2 is fairly old so you might want to nudge your management into moving ahead.

The "qla" means Qlogic which is ONE kind of fibre adapter you could have.

There are other types (e.g. Emulex).

So the first thing you'd need to know is which Fibre adapter you had.

I did some work with QLA and Clariion CX700s on RHEL 3 a few months back. There was a tool called something like "qla_scsi_scan" at Qlogic's site that you could download to do the rescan of SCSI.

Also - Are you running EMC PowerPath? EMC Navisphere? there would be additional commands to run on the host to make them see the new path(s).

Using PowerPath and Navisphere. I ran :

echo "scsi-qlascan" > proc/scsi/qla2300/2

you are supposed to get an entry in the file for the new device - but nothing changed

"powermt config" does not do anything for me

1) Have you verified you do in fact have a QLA card (as opposed to Emulex or something else)? If so how?

2) Does "lspci" show you the QLA card?

3) What does "powermt display dev=all" show?

4) Did you read what I said about the scan utility at Qlogic's site? The name of the utility is "ql-scan-lun.sh".

FWIW - Full sequence I used on RHEL 3 using QLA 2340 (QLA 2312 is what lspci displayed) was something like the following. As noted this was NOT RHEL 2 so may not work the same way even if you are using the same QLA card I was.

*************** DO THIS AT YOUR OWN RISK **********************

A) service PowerPath stop # Stop EMC PowerPath
B) service naviagent stop # Stop EMC Navisphere Agent
C) Renamed PowerPath configuration files in /etc:
emcp_devicesDB.dat, emcp_deviceDB.idx, powermt.custom
NOTE: mv NOT cp
D) ql-scan-lun.sh –s –r # Scan SCSI for new devices
E) service nagiagent start # Restart Navisphere
F) service PowerPath start # Restart PowerPath
G) powermt display dev=all # Show EMC device mappings

*************** DO ABOVE AT YOUR OWN RISK *********************

Note: The /dev/sd? designations will almost certainly change which shouldn't matter so long as you're using the PowerPath pseudo devices (e.g. emcpowera, emcpowerb etc...). The powermt display dev=all will show you which /dev/sd? paths go with which pseudo devices. On your next boot the /dev/sd? will likely change again as it did for me.

Already had verified it is a qla2312 and powermt ... shows me the 7 LUN that I already have mounted and no more

I have investigated your point of running the powermt stop start and nav agent start stop, but Qlogic recommend that you unmount any LUNS mounted and proceed with the start stop of above. As this is in production I would prefer to wait till the weekend and either do the above or reboot. Does the qla-scan-lun.sh not effect and LUNs already mounted ?

Thanks for you help here

When I did all the above I was actually migrating from one array to another (e.g. replacing the LUNs I was using with SANcopy LUNs of same from another CX700). Due to that I stopped everything on the system. Since it reorders your /dev/sd? devices that might be a good idea - looking at the details for the utility on Qlogic's site might help. I'm pretty sure that I ran it at least once with everything up with no untoward effects but wouldn't swear to it.

However, if you can stop your apps and/or db etc... then simply rebooting the box would probably be the simplest way to make it see the new device as it will automatically scan on reboot. I thought you didn't want to reboot though.

The systems I was doing were running Oracle RAC databases so it was a lot more complicated because there were two systems sharing the same LUNs. I basically extracted a subset of commands from the full plan I'd done for that.

Also as a reminder that was on RHEL3 systems - which a 2.4 kernel like RHEL2.1 so is probably very similar at the driver level but not necessarily exact. You'd of course want to have a good backup before you did any of this.

If you can get a downtime window then it might be worthwhile to try the scan without the boot. If it doesn't work then you do the boot but then in future you'd know one way or the other.

Don't feel bad though - coming up with the above even with EMC's help was like pulling teeth. We finally went to an offsite testing facility owned by Oracle with EMC contributed equipment to work it out and do the test.

For just adding a LUN rather than redeploying existing ones though I would think it shouldn't be that difficult. But as they say: Your mileage may vary.

Once we have Oracle down at the weekend, will try what you suggested - otherwise it is the old reboot :)

Thanks

I guess I should mention - you probably don't want to mv your EMC config files in /etc the way I had in above steps since you're just adding a LUN. In my scenario we were getting rid of the old LUNs so that is why we had to get the old config files out of the way.

P.S. Good Luck this weekend.