cancel
Showing results for 
Search instead for 
Did you mean: 

Windows 2003 Clustering ?

Former Member
0 Kudos

I'd be very pleased if a few of you who use windows 2003 clustering would give me a status report of how

it's been going.. any feedback welcome.

If we do it, we'll run windows 2003 64 bit, netweaver 7.0 and MII 12.0 on a certified ISCSI san.

Other than general impressions, I'd really appreciate feedback on the following items:

1) How long does it take to manually failover everything (including SQL DB) to the passive node?

I've read to expect up to 5 minutes.

2) If you pull the plug or something drastic, how long does it take to failover, and does it ever come

back 'dirty' ?

3) Has patching been a nightmare? I know that if something breaks the storage driver you are royally

screwed.. but other than that, after it's burned in, and if you use a QA environment - do you still cross your

fingers when you fail over and back?

4) Have you ever had trouble with resources refusing to go offline, or refusing to come online, for no

apparent reason?

5) I'm really interested in an answer to this question:

If one of the nodes in the cluster dies... do you have to build the replacement server right on the spot

and add to the same cluster, or could you build the node in another cluster (say the dev cluster) and then

add it to the remote production cluster (would require editing I'd imagine).

On this item, the thought really concerns me that we'd have to fly out a multidisciplinary team to the boondocks

if one of the cluster nodes fails. We have no remote IT support whatsoever.

Nobody in my company has ever built or supported a complex MSCS config, and I'm having a lot of trouble

finding answers to questions like these.. (experientially based - not technet/sapnote stuff).

Thanks.

Accepted Solutions (1)

Accepted Solutions (1)

Former Member
0 Kudos

1) How long does it take to manually failover everything (including SQL DB) to the passive node?

Depends. If you are talking ABAP only, then less than 5 minutes. If you are talking about ABAP + JAVA, then up to 15 minutes. The slowness here is due to SAP AS-JAVA, not the MS Cluster.

2) If you pull the plug or something drastic, how long does it take to failover, and does it ever come back 'dirty'

I have had a few server random reboots here and now, never a corrupted database. However, if you are using "mount points" on your storage, make sure you include the mount points as cluster resources.

3) Has patching been a nightmare? I know that if something breaks the storage driver you are royally screwed.. but other than that, after it's burned in, and if you use a QA environment - do you still cross your fingers when you fail over and back?

No problems w/ patching. We only patch security patches, we don't patch drivers unless we obsolutely need to. In my organization, drivers are bundled into what's called IPAKs, so instead of applying each individual drivers we go from IPAK 1 to 2, for example. Never had a problem w/ the cluster during IPAK upgrades.

4) Have you ever had trouble with resources refusing to go offline, or refusing to come online, for no apparent reason?

Had this issue before, but was due to the version of SAP supplied NT Cluster DLL files being out of date. Outside of this issue, no other problems.

5) I'm really interested in an answer to this question: If one of the nodes in the cluster dies... do you have to build the replacement server right on the spot and add to the same cluster, or could you build the node in another cluster (say the dev cluster) and then add it to the remote production cluster (would require editing I'd imagine).

Why would you build a replacement node somewhere else? If you are worried about the resources using the "unprepared" node, just configure the resource so it can only reside on the active node.

On this item, the thought really concerns me that we'd have We have Netweaver 2004s on Windows with SQL server.Now I am going to install ECC5 on the same server with installaing different database.Thus I would be having two SQL daltabase on same box.

Two, three, four, doesn't matter how many databases you have in your SQL server, as long as you have the memory for it, it's fine. To the SQL server, it's just a single engine hosting multiple databases. To SAP instances, they all talk to the same SQL server, but use different databases. Again, your biggest problem here will be system resources, not administration.

Now I want to know what are the preactions and post actions for it.As ports also for message server and all services will already be occupied,Then during installation which ports i should use or i can use any port number.Beside it on database level what additional I have to do.

SAP uses a different port range than SQL server. Keep your SAP instances on different ports, that's the biggest thing. The way I do it is, all production will use system number starting with 0. Test systems start with 1. Dev systems 2. Sandbox 3. So if I have multiple prod systems, their system numbers will be 01,02,03,04. If I have multiple test, then 11,12,13, etc.

Please suggest.

MSCS is not that difficult. I run it for my BI 7.0, GTS 7.0, E-Recruiting (ECC 6), and SAP R/3 Central Instances. Runs fine, I don't worry about it, downtime is minimal.

Former Member
0 Kudos

Excellent feedback. Just what I needed.

Since we will be running AS java, I can assure my comrades that clustering will NOT be "zero downtime".

-Kev

Former Member
0 Kudos

>

> Since we will be running AS java, I can assure my comrades that clustering will NOT be "zero downtime".

>

> -Kev

that's for sure. by default, ABAP only install the cluster resource defaults to 300 seconds for start up wait, before the resource reports an error. For our JAVA pieces, we upped this threshold to around 1200 seconds.

Answers (0)