PERC4/DC Clustering Question

Question

I originally posted this in the PowerVault forum but thought it might get more traffic here. < > Can someone point me to any docs here (or elsewhere) on configuring a cluster with the following hardware/software: 2x2850 with PERC4/DC (channel 0 for local HD, channel 1 for external array) 1x220s with 4x145GB HD Windows Server 2003 EE 1. 220s set to cluster 2. I have configured the drives and can boot node A to the OS and can see the drives in the 220s. 3. I can shut down node A and boot node B and see the drives on the 220s 4. If I boot node A into Windows and start node B, node A hangs and I get an audible warning I changed the configuration to cluster (writeback to writethru) but on subsequent boot the OS partition became unbootable (F1 to boot, F2 to setup). I've setup a cluster before but not recently - I'm sure I'm just missing a step or failing to set a parameter correctly. I seem to be generally on the right track (so it seems) if I could get the two servers to share the array. I seem to recall that we purchased a clustering kit for the last cluster I did 18 months ago but as I recall, it was primarily software and docs. Any pointers anyone? < > Addendum: After doing some searching, i came across a document I wrote 18 months ago on configuring the PERC cards that indicates that I may have 1) a SCSI ID conflict and 2) the PERC cards need to be configured for clustering (Management -> Objects -> Adapters) but when I did that it popped up a warning that it couldn't be set because one or more logical drives was in writeback mode. I'm close, but not there yet (wish I did this more often). greyseal53

FOURIER · Answer

hi,

if you have made sure that the SCSI id's are set correctly to 6 and 7 on your PERC adapters, changed the configuration to clustered in your ctrl-m bios, made sure that your SCSI cables are on the appropriate channels, etc...etc....I did not see you mention anything about setting your PV220 in clustered mode. You may want to ensure that this is in the right mode.

greyseal53 · Answer

I checked everything again last night and did not have the SCSI Initiator ID set correctly. Sorry for having failed to mention that the 220s' split-bus switch was indeed set to cluster.

So now, here's the setup - 220s in cluster mode; PERC adapter in node A is set to SCSI ID 7, write-thru enabled, cluster enabled, BIOS active; PERC adapter in node B is set to SCSI ID 6, write-thru enabled, cluster enabled, BIOS active.

When I boot each node individually, I can see the attached storage with no problem. The next problem to hurdle is overcoming the lack of a domain controller before I can build the cluster. I've found an MS Knowledge Base article that instructs how to run the cluster nodes as domain controllers. I'm off to study it in more depth...

Will report back my success with the cluster over the next few days.

greyseal53

FOURIER · Answer

Good deal!  Glad to know that your back in business.  One thing though, do NOT make one of your cluster nodes a domain controller, dhcp server, etc......that is a very big mistake, get some low budget server to do that function, do not put your data in jeopardy.

greyseal53 · Answer

Too late!!   I did indeed create a 'domainlet' using dcpromo on the cluster nodes. I wrestled with this for most of the day yesterday. I have the two 2850s with the PV220s and no budget for an outboard DC. In fact, to continue along the line of fault-tolerance, I saw a single outboard DC as another point of failure unless you cluster that as well and I really don't have the budget for 2 new boxes. I have an older Dimension 1.6GHz desktop sitting around and considered using that, but given its age and the fact of space constraints where I'm locating the rack, led me back to using a domainlet with the cluster.   Got all my other problems tackled last night (BIOS ID corect, both nodes can access the array) but ran into another.   Question: I have a 4x145GB 10K array set up as two spans with two disks each. However, all the MS docs recommend that the quorum drive be on a seperate physical spindle(s). I presume I can mix drive sizes within the 220s (say if I acquire 2 36GB 10K spindles)? If so, since I plan to set the quorum size at 1GB (twice the recommended size) what happens to the other ~34 GB of space - waste? Everything I read says don't put any data on a partition on that spindle - that seems silly.   I guess I'll be hitting the Dell reps up today for two more drives...   greyseal53

greyseal53 · Answer

Here's my config (for anyone who wants to chime in):

I have 2 2850s each with PERC4/DC cards. Channel 0 is handling the internal drives (2x36GB) and channel 1 is handling the outboard PV220s (same for each 2850).

The PV220s currently has 4x146GB main drive array. I plan to add 2x18GB drives for quorum tomorrow. The 220s has the split-bus switch set to cluster. Each PERC card is set to cluster and WRITETHRU and the SCSI IDs are set to 6 & 7 on each PERC respectively. However PERC BIOS is enabled - if I disable it, the internal drives on the 2850s cannot be located on boot (no OS). A note here - I read on a thread on Google that the internal drives need to be on their own controller card and not on the PERC with the PV220s - is that truly the case?

On my first pass with the 4x146GB array set as RAID 10 drive (2 spans), I was able to boot to Windows Server 2003 on node A and it sees the PV220s drives. I set up two logical drives one for the quorum (Q:) and the rest for storage (S:). I could boot up node B and it can see the PV220s as well. I shut down node B and attempted to build the cluster using node A but it fails when it attempts to locate a quorum drive (uses local resource).

I've ordered two 18GB drives to create a seperate physical quorum span and will test this tomorrow, but for now I have some concerns on this setup. Any suggestions (particularly from Dell)?

greyseal53

PowerEdge HDD/SCSI/RAID

PERC4/DC Clustering Question

Was this post helpful?