HPE 9000 and HPE e3000 Servers
1748085 Members
5112 Online
108758 Solutions
New Discussion

RP3410s and SCSI disk trays

 
SOLVED
Go to solution
rventura
Frequent Advisor

RP3410s and SCSI disk trays

Hello,

 

I have 2 HP RP3410 servers. They run our ERP system which data is in a ServiceGuard package and stored in 2 HP Disk System 2120. Each of the disk trays has 2 SCSI connectors and each connector goes to a different server. This is pretty standard kind of thing. Each try has 2 HDDs and they are mirrored.

 

Now, due to new requirements, i need to break the cluster and completely remove 1 of servers. This means that i will be disconnecting the server from the disk trays, leaving an open SCSI connection in each tray.

 

My concern is that unplugging the SCSI cables could break something... I really havent done a lot of work with SCSI, but i have heard that they need to be terminated...? Is this still true..?

 

Can anyone provide some help on the matter?.

 

Thanks.

3 REPLIES 3
Torsten.
Acclaimed Contributor

Re: RP3410s and SCSI disk trays

You need to plug a terminator on the open port of the ds2120.


Hope this helps!
Regards
Torsten.

__________________________________________________
There are only 10 types of people in the world -
those who understand binary, and those who don't.

__________________________________________________
No support by private messages. Please ask the forum!

If you feel this was helpful please click the KUDOS! thumb below!   
Matti_Kurkela
Honored Contributor
Solution

Re: RP3410s and SCSI disk trays

If all the servers and disk trays are properly grounded, disconnecting the cable should not cause any hardware damage even if the system is running. But you may see some SCSI bus errors as a result of the disconnection.

 

Serviceguard A.11.19 documentation still mentions in-line terminators for old Fast/WideSCSI-based two-node cluster setups. Basically, a in-line terminator is a short length of SCSI cable with the terminator pack embedded into one of the connectors.

 

The cabling should be like this for each SCSI bus if in-line terminators (ILT for short) are used:

node1 <--ILT--><SCSI cable> storage <SCSI cable><--ILT--> node2

If a cluster node needs to be disconnected while the other node is running (e.g. for hardware service/replacement purposes), you're supposed to disconnect the cable from the node side only, and leave the ILT and the SCSI cable attached to the storage.

 

If the cluster is halted when you're making the changes, disconnecting the cable from the storage system and putting on a SCSI terminator instead is fine.

 

For modern LVD SCSI, I'm not sure if equivalent ILTs exist. The QuickSpecs document for ds2120 indicates that there are "self-terminating SCSI cables" (product C7541A and C7520A), which I guess are basically SCSI cables with ILT built in.

MK
rventura
Frequent Advisor

Re: RP3410s and SCSI disk trays

Ok, so i will just leave the cables connected to the D2120 and unplug them from the server. Once we can schedule some brief downtime, i will halt the package, unplug the cables from the ds2120 and plug in the terminators.

 

Thanks for the help.