Google+

Merging Fiber Channel Fabrics

December 3rd, 2008 No comments

You have two fabrics running off of two switches. You’d like to make them one fabric. How to do that? For the most part, it’s simply connecting the two switches via e_ports.

Before doing that, however, realize there’s several factors that can prevent them from merging

1. Incompatible operating parameters such as RA_TOV and ED_TOV
2. Duplicate domain IDs.
3. Incompatible zoning configurations
4. No principal switch (priority set to 255 on all switches)
5. No response from the switch (hello sent every 30 seconds)

To avoid the issues above:

1. Check IPs on all Service Processors and switches; deconflict as necessary.
2. Ensure that all switches have unique domain ids.
3. Ensure that operating parameters are the same.
4. Ensure there aren’t any zoning conflicts in the fabric (port zones, etc).

Once that’s done:

1. Physically link the switches
2. View the active zone set to ensure the merge happens.
3. Save the active zone set
4. Activate the new zone set.

EMC SRDF Basics

December 3rd, 2008 No comments

Conceptually and operationally, SRDF is designed to work in a WAN/Internet/Cloud/SAN environment with multiple Symms involved, while Timefinder is local to a Symm, but performs the same functions.

The difference, SRDF can be performed without Geographic boundaries, while Timefinder is local. The following are various different forms of SRDF that can be used by a customer to perform SRDF operations. 

Synchronous mode

With Synchronous mode, the remote symm must have I/O in cache before the application receives the acknowledgement. Depending on distance where these Symmetrix machines are located, this may have a significant impact on performance. This form of SRDF is suggested to be implemented in a campus environment.

If you want to ensure that the data is replicated real time without dirty tracks from one symmetrix to the other, you might want to enable Domino effect. With Domino effect, your R1 devices will become not ready if the R2 devices cant be reached. 

Semi-synchronous mode

With Semi-synchronous mode, the I/O between the R1 and R2 devices are always out of sync. The application receives the acknowledgement from the first write I/O to the local cache. The second I/O isn’t acknowledged until the first is in the remote cache. This form of SRDF is faster than the previous mentioned Synchronous mode. 

Adaptive Copy-Write Pending

With Adaptive Copy-Write Pending, all the R2 volumes are copied over without the delay of acknowledgement from the application. With this mode, we can setup a skew parameter that will allow max number of dirty tracks. Once that number is reached, the system switches to a preconfigured mode like the semi-synchronous mode until the remote data is all synced. Once this is hit, SRDF is switched back to Adaptive Copy-Write Pending mode. 

Volume Logix

December 3rd, 2008 No comments

The order for getting fibre channel based hypervolume extentions (HVEs) viewable on systems, particularly SUN systems, is as follows:

1. Appropriately zone so the Host Bus Adapter (HBA) can see the EMC Fibre Adapter (FA).

2. Reboot the system so it can see the vcm database disk on the FA OR
1. SUN:
1. drvconfig -i sd; disks; devlinks (SunOS <= 5.7)
2. devfsadm -i sd (SunOS >= 5.7 (w/patches))

2. HP:
1. ioscan -f # Note the new hw address
2. insf -e -H ${hw}

3. Execute vcmfind to ensure the system sees the Volume Logix database.

4. ID mapped informationi
1. Map HVEs to the FA if not already done.
2. symdev list -SA ${fa} to see what’s mapped.
3. symdev show ${dev} to ID the lun that ${dev} is mapped as. The display should look something like:

Front Director Paths (4):

{
———————————————————————————————–
POWERPATH DIRECTOR PORT
———————- —————– ————
PdevName Type Num Type Num Sts VBUS TID LUN
———————————————————————————————–
Not Visible N/A 03A FA 0 RW 000 00 70

Not Visible N/A 14A FA 0 NR 000 00 70

Not Visible N/A 03B FA 0 NR 000 00 70

Not Visible N/A 14B FA 0 NR 000 00 70

}

The number you’re looking for is under the column LUN. Remember, it’s HEX, so the lun that’ll show up on the ctd is (0x70=112) c#t#d112

5. On SUN systems, modify the /kernel/drv/sd.conf file so the system will see the new disks. You’ll need to do a reconfig reboot after modifying this file. If the system doesn’t see it on a reconfig reboot, this file is probably the culprit!

6. fpath adddev -w ${hba_wwn} -f ${fa} -r “${list_of_EMC_devs}”

You can specify multiple EMC device ranges; just separate them by spaces, not commas
7. Reboot the system so it can see the new disks on the FA OR
1. SUN:
1. drvconfig -i sd; disks; devlinks (SunOS <= 5.7)
2. devfsadm -i sd (SunOS >= 5.7 (w/patches))

2. HP:
1. ioscan -f # Note the new hw address
2. insf -e -H ${hw}

SUN Host, EMC SAN, how to make it work?

December 2nd, 2008 No comments

EMC disks use luns vs scsi target numbers. SUN doesn’t normally use lun numbers so, to make the two talk together, you have to edit the /kernel/drv/sd.conf file as follows:

Normal, unedited file

name=”sd” class=”scsi”

target=0 lun=0;

name=”sd” class=”scsi”

target=1 lun=0;

Edited file

name=”sd” class=”scsi”

target=0 lun=0;

name=”sd” class=”scsi”

target=0 lun=1;

name=”sd” class=”scsi”

target=0 lun=2;

name=”sd” class=”scsi”

target=0 lun=8;

name=”sd” class=”scsi”

target=1 lun=0;

name=”sd” class=”scsi”

target=1 lun=1;

name=”sd” class=”scsi”

target=1 lun=2;

name=”sd” class=”scsi”

target=1 lun=8;

Add the luns for each of the EMC targets. Limit the targets to the EMC disks only, however, as the extra luns will dramatically increase the boot times on the system.

If you’re using more than 9 luns, make sure you use the hexadecimal notation for the lun numbers:

lun=0xa;
lun=0xb;
lunc=0xc;

and ensure each lun line ends in a semicolon.