Fibrechannel general questions

[xrated]

Is it allowed or is fibrechannel concipated that i can change the wiring when the devices are still on? (unplug cables, change fibrechannel switch, replace cards and so on)

I did that today and when i repluged the cable on a specific device it only had a link after i turned off/on the server. The other devices were no problem.

Plus, i have a RA4100 storage with 2 controllers and on one controller it shows no heartbeat, nothing. Have i destroyed something by just plugging the cables?

Does anybody know how to control active/standby card with cpqfc.ham ? I do not know which card is active! But i have to because i want to replace one fibrechannel hub with a switch and the other hub is just for rescue cases

 

[trumpeterman2000]

It is designed and possible to hotplug the cables during production. Usually you will have redundand pathways anyway (either selected storage presentation or securepath). But you have to make sure the FC-Hub or FC-switch is not switched off, as there is a special boot sequence to be proceeded, once the hub or switch has been siwtched off.

Secondly, once you changed the hba or another part of the SAN, you will need to reconfigure your system. Every part of the SAN is using unique

http://WWW-names

(comparable with MAC-addresses on NIC's) For instance if you are using zones within the fabric, you will need to tell those zzones the new

http://www-name,

as well as to the allowed connections to the LUNS etc. Just check, in which settings you might have to set/use a

http://www-name,

each of those settings need to be altered after replacing fc-hardware.

Further you need to check on your zones anyway, there is hard zoning and softzoning. If you used hard zoning, the zones depend just on the used port. Once you used another port on the switch while using hardzoning, you might not see the hba / lun etc. You will be in trouble, if you didn't have a documentation before and after the change.

pretty much same goes for the controller in the ra 4100, as it is a fc device with

http://www-name.

Besides, one common failure with fibre optic cables and gbic's is to either let them touch the ground, which will put dust on to the fibre, or even to touch it. first one will just knowck down your performance, if you are lucky, the second one prety much makes the cable or gbic useles because of the skin grease, which now is on the fibre either.

 

[xrated]

OK, we do not have a software like securepath but the driver for the HBA (cpqfc.ham under Netware) switches automatically to the other HBA if something fails

We haven´t got any zones, the hub isn´t even to manage do i need to setup zones when we only have Netware?

Do you whats wrong with the controller in the RA4100, i do not understand by it is not working anymore, somebody said it might be deactivated because of link was gone

 

[trumpeterman2000]

Okay, under netware does the driver the switching al right. If you use netware only, there should not be much trouble without zoning. As soon, as another OS comes into the game, you will need to have zones, otherwhise you might crash your luns.

Hoever, I learned in your first posting, that you changed some hardware, i.e. an HBA. Usually you have to configure in the ACU (Array Configuration Utility), which hba is allowed to access which lun. even if you are running netware only, usually you don't want to have several servers to have access to one lun. This hba is "named" using his

http://www-name,

which is unique to each hba. So within ACU you give access to the lun to hba with

http://www-name

xx.xx.xx.xx.xx.xx. (where x stands for a hex-code). However, after changing that hba ( and therefore the

http://www-name)

you still have to tell the RA 4100 controller, that a new player is in town, menas you need to reconfigure the access for the new

http://www-name.

Unless you did that already, indeed the link has gone for the controller, as it doesn't see "his" configured hba anymore.

How exactly does it show to you, that the controller is not working anymore? Do you get any errors in SWCC or ACU?

Do you have any link led's on the hba / hub / ra4100 controller?

 

[trumpeterman2000]

Had another ,look on that problem. Since you don't have secure path installed, the driver itself is controlling only the pathways. You pulled the cable, therefore the driver switched over, and by default it obviously does not switch back when the link becomes alive again. So your controller indeed thinks, his path is dead, and does not do anything.

Have a look here, it will tell you most of the things you need to know about the RA4100: ftp://ftp.compaq.com/pub/products/storageworks/RA4100/14U2-0401A-WWEN.pdf

 

[xrated]

thanks a lot!
I think Netware handles it a bit different. In ACU there are no options to configure a lun or

http://www names

Maybe these options only show up when secure path is installed or our hardware does not support manual luns or

http://www names

 

[xrated]

Now í´ve shutdowned everything and disconnected one Hub and connected a Switch in place. (We had 2 Hub´s before)

Strange thing was that one connection from storage to the switch was up 10-20 minutes after it had a link!

Therefore we have no Secure Path installed i cannot see which HBA´s are active, so i don´t know the installed switch is active or standby

 

[xrated]

Other thing that is irritating me that the port status LED´s one some ports (0,2,4) are switching to off sometimes
The ports 1,3 are always on

It´s an Compaq 2/8EL Switch
All is working despite the not active port status LED´s

 

[trumpeterman2000]

Hi there, sorry for the delay .

Just to make the situation a bit more clear: Which version of netware is used with which patch level, and which driver version is used either? How much servers are attached to the SAN? What kind of HBA's? (probably Emulex, but which ones?) Are you aware, that not every version of netware is working here properly, hence not supported by CPQ? (ftp://ftp.compaq.com/pub/products/storageworks/ra4100/RA4100-SANSupportMatrix.pdf)
The 2/8 EL switch supports up to 2 GB's /s, but you will need to have different GBICS for that. However, did you set the switch to auto negotiation, or to a fixed value?

Did you make sure to have the right boot sequence (first siwtch, then primary storage => backup units and servers last? The switch needs to be fully initialised before switching on the other parts. Otherwhise you might experience situations like having the connection very late or even not ready at all, even if a physical link is established.

You said you were using hubs in first place before upgrading with that switch. Those hubs very likely were FC_AL hubs? (arbitrated loop)? I doubt you can have one pathway running over FC-AL and the other Pathway via FC-SW (switched fabric), unless you configured the used ports on the 2/8 EL switch to be Loop-ports (FL ports). But then you would not need the upgrade to 2/8 EL switch at all, unless you have other servers in the SAN and quite a good zoning concept.

Have a look on the switch itself, you will see how the ports are configured, and see the login history.

In the end of the day I doubt either, that you can "manage" the cpqfc.ham driver in the way you maybe want to. Tharefore they invented securepath, which is controlling and managing the pathes. The driver itself very likely just sits and watches the links.


Best way to see the whole configuration of the switch is "supportshow" , it really shows everything.

 

[xrated]

There are 3 servers attached, all Compaq Proliants with Netware 5.1SP4, Cluster Services SP3

cpqfc.ham is version 2.52 (NSSD should be version 6.40A i think)

HBA is also from compaq, should be model 3200 if i remember right

Yesterday i had connected the fc hub and the new switch, it all worked properly besides the failback which is due to the missing secure path software. Btw do you know how much the software is?
One thing that was irritating me that one storage had an active link to the switch and the other hub had an active link to the hub!! How can that be, the servers can´t access to hubs or switches at once!
So today i just disconnected the hub to make sure the switch is working alone
Is that what you mean with boot sequence? Where can i set up a boot sequence? The hubs can´t be managed
I am missing a little of theority with this things

With this loops, the only thing i know is that a fabric can be compatible with older devices that are not fabric enabled, if looping is enabled. Looping is not enabled on the switch (i am not sure), i think the license for that is missing too

what is arbitrated loop and el ?
the port speed is set to auto on the swith but seems to work
traffic on the switch is lower that i thought, like only 1MB/s max. 3-7MB/s

 

[connollyg]

I think you need to get a expert in (ie spend some money)someone who can advise you on what is wrong, and also more importantly what is supported in your environment and what your options are!

G

 

[xrated]

i think not, thanks anyway

there is nothing wrong i just want information

 

[trumpeterman2000]

sorry for the delay, I was not in office for a few days.

Boot sequence: the order you switch on your SAN. you need ALWAYS the following order: FC-switch(es). It takes a few minutes, until they have ran all selftests and are conpletely initialised. Then you can switch on your primary storage (RA 4100), afterwards the Backup units (libraries etc.) and in the very end the servers. The whole process might take easily 15 or twenty minutes! Otherwhise the devices can't logon to the switch properly, and you will see failed links and such stuff.

Will be back tomorrow

 

[xrated]

Thx!
i saw in the manual that the switch has a boot time of up to 7minutes

 

[trumpeterman2000]

the 3200 very likely stands for the array controller, which (Smart Array Controller 3200).
The thing I not understand is, that you are using a FC hub as well as a FC switch, they are using different topologies. The hub is using (FC-AL)arbitrated loop, an older technology, and the switch is using switched fabrik (FC-SW). It should not be be possible to use both of them at one time. You might be able to cascade one FC-AL hub to a FC-SW switch, but thats a completely different matter.

If those hubs are unmanaged, they must be quite old, aren't they? Anyhow the switch can be mnaged via serial connection, telnet or web gui. the last one is actually quite comfortable

Usually you would not enable the whole switch to be looping, but just one or a few ports to be L port (get more informations about terms and topology on the net, i.e.

http://www.attotech.com/fcgloss.html

where exactly do you measure the speed? Fibre channel technology is capable up to 2 Gb/s, so I doubt your data really are transmitted with 1-7 MB/s. If yes, something really is messed up

Btw, why don't you just call the Cpq-hotline, as your switch should be within warranty. They should be able to help you, most of those guys are quite okay, unless you come out on some outsourced line, like Stream

 

[xrated]

forgotten this thread a bit ;-)

We had this switch only for testing speeds. We had even 2 switches but not sufficient cables for it so i used only one switch and the other hub as standby.

For us the matter is more the server itself, it has sometimes very high cpu usage so thats the bottleneck

I have measured the speed when i copied files on the server. When files are copied on 2 workstations at once then the server isn´t able to deliver it as fast as possible because the cpu usage goes up to 100%.
Thats the problem i always say but i was forced by my chief to test a faster switch ;-)