AIX's mpio 1

[mag007]

I am in the process of setting up MPIO on AIX 5.3; I already have 2 HBAs, but I am not sure if anything needs to be configured on the switch side to enable MPIO.

The AIX filesets are there, however I am only able to get 1 path active, eventhough I have 2 HBAs.

lspath
Enabled hdisk0  scsi3
Enabled hdisk1  scsi3
Enabled hdisk2  fscsi0
Enabled hdisk3  fscsi0
Enabled hdisk4  fscsi0
Enabled hdisk5  fscsi0
Enabled hdisk6  fscsi0
Enabled hdisk7  fscsi0
Enabled hdisk8  fscsi0
Enabled hdisk9  fscsi0
Enabled hdisk10 fscsi0
Enabled hdisk11 fscsi0

What about fscsi1 ?

 

[p5wizard]

Are the LUNs on the SAN server assigned/presented to both adapter's WWNNs? Is your zoning info on the switches up to date?

In any case, with standard MPIO, you would see both paths with lspath, but the IO would be routed to one of the paths, with failover to the next path only. You need other PCMs to enable load-balancing also. E.g. sddpcm for IBM DS8000/6000 or ESS LUNs.


HTH,

p5wizard

 

[mag007]

Okay, so I should probally speak with my SAN team to verify if they are presenting LUNs to both HBAs.

 

[kHz]

It is on the SAN team's side.

 

[mag007]

khz:
thanks.

Here is what I have now:

lspath
Enabled hdisk0  scsi3
Enabled hdisk1  scsi3
Enabled hdisk2  fscsi0
Enabled hdisk3  fscsi0
Enabled hdisk4  fscsi0
Enabled hdisk5  fscsi0
Enabled hdisk6  fscsi0
Enabled hdisk7  fscsi0
Enabled hdisk8  fscsi0
Enabled hdisk9  fscsi0
Enabled hdisk10 fscsi0
Enabled hdisk11 fscsi0
Enabled hdisk2  fscsi1
Enabled hdisk3  fscsi1
Enabled hdisk4  fscsi1
Enabled hdisk5  fscsi1
Enabled hdisk6  fscsi1
Enabled hdisk7  fscsi1
Enabled hdisk8  fscsi1
Enabled hdisk9  fscsi1
Enabled hdisk10 fscsi1
Enabled hdisk11 fscsi1

Is it safe to say I have multiple paths now?

 

[kHz]

Yes. But why are the two paths on separate adapters? You have ony single-channel hba's?

 

[p5wizard]

What you also need to do is verify if your healthcheck interval is set to non-zero, so that paths which failed (switch down, port failure, cable unplugged, whatever) and are then made available again, will reintegrate in the MPIO failover scenario.

So, if the disks are not in use (VGs varied off):

for i in 2 3 4 5 6 7 8 9 10 11
do
chdev -a hcheck_interval=60 -l hdisk$i
done

Or if the disks are in use:

for i in 2 3 4 5 6 7 8 9 10 11
do
chdev -a hcheck_interval=60 -l hdisk$i -P
done

In which case you need a reboot to make your changes active.

As for the failover to take place, I'd run a test scenario on a non-productive server/partition first and see what happens when you unplug one of the fiber cables or "unpresent" a LUN from the first WWNN of this test server on the SAN side.



HTH,

p5wizard

 

[mag007]

Yes.  But why are the two paths on separate adapters?  You have ony single-channel hba's?

I have 2 HBAs.

I am not sure what p5wizard says... :-(

 

[p5wizard]

normal situation:

lspath|grep hdisk2
Enabled hdisk2 fscsi0
Enabled hdisk2 fscsi1

you pull fscsi0 cable:

lspath|grep hdisk2
Failed hdisk2 fscsi0
Enabled hdisk2 fscsi1

you still have access to hdisk2 - failed over to fscsi1 - no problem.

you replace fscsi0 cable:

lspath|grep hdisk2
Failed hdisk2 fscsi0
Enabled hdisk2 fscsi1

The paths to the disks via fscsi0 are restored, however they stay "Failed", until you manually set them to "Enabled" with chpath command. You would not have failover capability until you set the Failed paths back to Enabled (if now something would happen to fscsi1 also...)

This is the default situation with hcheck_interval=0 (see lsattr -E -l hdisk2 output)


If you have hcheck_interval set to 60, the paths become "Enabled" again a short while after the path is physically restored.



HTH,

p5wizard

 

[p5wizard]

This is if you do not have any vendor-specific path control module and use the standard AIX path control module. If you have a vendor-specific PCM, then please ignore all I said (in this post ;-))

HTH,

p5wizard

 

[mag007]

p5wiz:

Thanks for the explanation! When I get on site, I will try the test. For some reason, its tough to get documention on this MPIO. I think, I like this better than EMC's powerpath.

 

[p5wizard]

Well AIX-MPIO with standard PCM does not provide load balancing, only failover mode, whereas EMC's powerpath provides both I believe.

If you can live with that, it's OK, because it has this BIG plus: on AIX upgrade, no need to do special pre- or postprocessing (re vendor specific multipathing software). It all comes standard with standard AIX.


HTH,

p5wizard

 

[mag007]

Can you have both, PowerPath + AIX's MPIO installed?

 

[p5wizard]

I don't have experience with powerpath, but IMHO it is either MPIO or a vendor specific driver (MPIO or IBM-SDD, MPIO or PowerPath, ...).

However, there may be a vendor specific PathControlModule for EMC, so that you use AIX's framework MPIO with a special PCM for EMC, which may or may not (I don't know) provide load balancing AND path failover.

This is how we use MPIO for IBM DS8000 LUNs: by installing SDDPCM, which provides the SDD functionality in an MPIO framework. And that is one way to accomodate LUNs of different vendors on one SAN-client, as it is possible to install multiple PCMs on top of AIX MPIO.


HTH,

p5wizard

 

[kHz]

You can't have both.

MPIO has both failover and load balancing just like Powerpath.

 

[p5wizard]

MPIO has loadbalancing if the PathControlModule being used supports it. The PCM which is native to AIX doesn't. It is provided so that you have a path-failover solution using MPIO if you don't have/buy a vendor specific solution. That way at least you have a failover system...

For MPIO with load-balancing, you need to acquire a vendor-specific path-control-module for your specific SAN box. E.g. SDDPCM implements SDD functionality (load-balancing and path failover) for IBM ESS/DS as a PCM for MPIO/AIX.



HTH,

p5wizard

 

[mag007]

This is how we use MPIO for IBM DS8000 LUNs: by installing SDDPCM, which provides the SDD functionality in an MPIO framework. And that is one way to accomodate LUNs of different vendors on one SAN-client, as it is possible to install multiple PCMs on top of AIX MPIO.

Can you please add a little bit more acronyms? ;-)
You must be an ex-ibmer

Thanks again for all the help. Leared a lot about MPIO.

 

[mrn]

No if he was ex-IBM he wouldn't know what he is talking about...

Mike

"Whenever I dwell for any length of time on my own shortcomings, they gradually begin to seem mild, harmless, rather engaging little things, not at all like the staring defects in other people's characters."

 

[p5wizard]

Close...

DS8000 is an IBM SAN storage server (follow-on to ESS or Enterprise Storage Server). DS6000 midrange and DS4000 entry models exist also - DS4000 is what used to be FastT.
IBM has DS4/6/8000
EMC has Symmetrix, Clariion
HP has EVA4/5/6/8000
Hitachi has (don't remember, sorry)

MPIO = MultiPathIO

PCM is already explained: PathControlModule, necessary for MPIO to function

SDDPCM is a PCM provided by IBM for use with DS6000 DS8000 ESS in an MPIO scenario (load balancing and path failover)

AIXPCM is a PCM provided by IBM, comes with AIX (only path failover)

LUN = SAN volume

I thought all TLAs and ETLAs were explained in my posts prior to using 'em...

Oh and:
TLA is a ThreeLetterAccronym
ETLA is an Extended TLA (TLA with more than 3 letters) ;-)



HTH,

p5wizard