HSZ80 Config interpretation

[Srivathsan]

Hi All,

Could anybody please help me interpret the attached show conn full output on my HSZ80 in Multibus Failover config.

I'd be glad to provide with more data if required.

Thanks in advance

Srivathsan

[Uwe Zessin]

Srivathsan,
I think you have a storage system with two HSG80 controllers, not HSZ80.

There are two fibre channel adapters involved - they have the WWNs 1000-0000-C922-1500 and 1000-0000-C925-95D8. It is not clear to me if this is one server or two servers involved.

I guess the configuration has a fibre channel switch from Brocade with the core PID addressing turned off. Both adapters and all four ports of the HSG go to a single fibre channel switch with domain-ID:1.

Adapter 1000-0000-C922-1500 goes to port 0 of the switch, the other one goes to port 1.

Each adapter has 4 connections with unit offset 0, so the controllers are running in a MULTIBUS_FAILOVER configuration, but you wrote the latter already. Offset 0 means that each adapter can potentially access all units on the HSG unless selective storage presentation is used on the units.

The OPERATING_SYSTEM value and the fact that the controllers are running in MB_FAILOVER suggest that there is Tru64 Unix version 5 involved.

I hope that was somewhat helpful. If not, please ask.

[Srivathsan]

Thanks for your reply..

But I'd actually want to learn more.

Please find attached the config details in more detail.

Would appreciate anything that you people might want to add on this.

Thanks in advance
Srivathsan

[Uwe Zessin · Accepted solution]

OK, so from the picture you have two servers, each with one fibre channel adapter. And you have two HSG80-based subsystems, each with two HSG80 controller modules.

According to the report you have provided, the cache batteries for the storage system TARIFA2/DC2 will expire on 22 June 2004. If they have not been replaced since this report was generated, you should schedule a replacement next month. These batteries are only good for two years after which they are no longer good enough.

I had edited your first report a little to remove an artificial line break. It looks like I didn't do that properly. I now see that node TARIFA2 has a unit offset of 50 on the connections. That means that node TARIFA can potentially access all units, but TARIFA2 cannot access units D0 through D49. On the other hand I see that the controllers are running in SCSI-2 mode. If I recall correctly, this allows a fibre channel adapter to only see the first 8 units, but I am not totally sure - it's too long since I have worked with this.

It looks like sombody has built a cheap desaster-tolerant configuration - don't get me wrong, I don't mean to be insulting. What do you want to do? Have you become the administrator of this configuration?

[Srivathsan]

Thanks very much Uwe.

Yeah you are right about my concern on this setup and the reason for trying to understand it. ( I am the administrator !!)

Thanks for your info, I'll have the battery replaced before it expires.

Just wondering the effect otherwise ( I believe write back caching is disabled right ? )

Thanks
Srivathsan

[Tarmo Linnam?gi]

Hey

Seems like you have two hosts with one fibre channel adapter.
You can see this on the host ID (there is two different ID-s) and on the adapter ID (one ID per host).
First adapter has offset 0 and second adapter has offset 50.
What else do you need?

BR

Tarmo

[Uwe Zessin]

Hello Srivathsan,
write-back caching can be selected individually on each unit and it is enabled by default (a "SHOW UNITS FULL" command will reveal this). If the battery fails, the controller will disable access to the units.


Tarmo,
it depends on the host's drivers if he assigns a common WWNN to all adapters in the box. I will attach an example. It is a single Tru64 Unix system with two adapters - as you can see the WWNN is different although the adapters are in the same box.

[Tarmo Linnam?gi]

Uwe

True - but in this kind of situation im thinking about two hosts because the difference on offsets - don't think its reasonable to have different offsets on 2 adapters per machine. You can ofcourse but...can't see the need on MultiBus

BR

Tarmo

[Srivathsan]

Hi Folks,

This sounds a bit confusing to me.

Would you want me to provide you with some more command outputs.

I'd want to understand my setup in detailand would appreciate any help in this regard.

Thanks
Srivathsan

[Uwe Zessin]

Tarmo,
I agree with you. The different offset suggest different servers, but I have learned to be very picky.


Srivathsan,
I use a set of commands to fetch the current configuration - I'll attach it. If you can post back the output I am sure me and others are happy to rip it apart.

Do you know what fibre channel switches you have in use and if you have command line access to them? There could also be 'interesting' information hidden in them.

[Srivathsan]

May 26, 2004 09:44:51 GMT

--------------------------------------------------------------------------------
Ok..thanks...here i go..

There are 2 hosts TARIFA and TARIFA2 in 2 different DCs.

There are 2 switches ( compaq SAN Switch ), not sure if i can manage that, 1 in each DC.

The outputs you had asked for are attached and are gathered from the syscheck run on TARIFA a few days back ( there's been no config change in the hosts after that )

The HSG80 storage subsystem in each DC has 2 Controllers in Multibus Failover mode.

4 Fiber cables from each HSG80 Storage subsystem ( 2 Fiber cables from each controller )run into the local Compaq SANSwitch.

One fiber cable runs from the SERVER's Fiber Adapter to the SANSWITCH

One fiber connection runs between the 2 SANSwitches connecting the DCs

The HSG80 Storage subsystem in each DC provides HARDWARE RAID5 LUNs ( 11 in number )

These eleven luns can be broken into 6 luns + 5 luns.

Six luns from each HSG80 storage subsystem, 6 local + 6 from remote DCs HSG80 are used by TARIFA

Similarly the remaining 5 LUNs on each site are used by TARIFA2.

And these 6 and 5 LUNS are mirrored to their remote counterparts using LSM.

I hope am making sense.

Thanks
Srivathsan

[Uwe Zessin]

Hello Srivathsan,

your analysis looks pretty thorough to me - let my try to fill in some blanks:

- a SANswitch can be managed. It depends on the type of switch what management paths you have available.

How many ports does the switch have? Does it have a small display and what color is the display? Does the switch have SC-style connectors?

- your storage systems have a mixture of RAID-1 (e.g. Unit D0), RAID-5 (e.g. D21) and RAID 0+1 (e.g. D101)

- your storage system is based on the Modular Array, not the older StorageWorks I

- it looks like the system has been 'de-tuned' a bit, because the MAX_READ_CACHED_TRANSFER_SIZE and MAX_WRITE_CACHED_TRANSFER_SIZE for most units have been set pretty low.

- there are no preferred paths set on the units - all units are currently online to one controller and the other is unused

[Srivathsan]

Thanks Uwe,

Any and all tuning suggestions are welcome.

Srivathsan

[Uwe Zessin]

Srivathsan,
there might be a valid reason for this change that I am not aware - anything in the documentation?

[Dave Pinsker]

Uwe:

> If the battery fails, the controller will disable access to the units.

Not sure where my info came from - may have been a Compaq/HP support Gold TAM - but I have been led to believe that if a write-back cache battery fails it will disable only the cache, not the units themselves.

Which information is correct?

Thanks,

_Dave_

[Uwe Zessin]

Dave,

the correct answer, as usual, is 'it depends' and one (including me) should never generalize ;-). There is a nice table in the HSG80 trouble shooting reference guide. It depends on the type of storage set, mirrored or non-mirrored caches and one or both batteries failing.

E.g. RAID(5)sets and MirrorSets become inoperative when both batteries have failed. Both controller continue to function normally if one battery has failed, but the other is more than 50% charged.