MIMIX Recommendations for 8.12

We are setting up MIMIX next week.
I am wondering if we should use OWJRNL for MIMIX or should we create a new JRNL just for MIMIX?
Any advice or tips would be greatly appreciated.

We have been using Mimix for many years and have always used the OWJRNL for all the JDE business data as is normally used by JDE. We created a separate journal library for everything else with multiple journals and associated journal receivers. For example, all the Printqueue files are associated with it's own journal, package libraries and IFS directories have their own journal, etc.

We use Mimix with JDE. I think you must use the OWJRNL definition as defined in your jde.ini as JDE will automatically start journalling for tables when an application is configured for transaction processing. We found a few items that have helped our implementation:
1. If on XE/ERP8, put the PRINTQUEUE file in a separate journal/data group. Never do a sync with the *SAVRST option on this file as it will lock the file and all UBEs will lockup trying to create a new member.
2. Put all IFS entries in a separate journal/data group.
3. Exclude the /sys_dir/LOCALE/xml directory from the journal/data group. JDE requires exclusive access to the license files in this directory and services will not startup if journalling is enabled.

I'd use OWJRNL. If you create a new one, and end up with tables in the same Environment with different journals, you will start getting commitment control errors.

Just use OWJRNL. We have been on Mimix here for years.

JMHO
Tom

Is there a limit to the number of files you can journal?
There is concern of using OWJRNL for the MIMIX journal and having a number of files in the journal only for MIMIX.

Will journaling effect performance if all JDE files are journaled?

Can MIMIX have mulitple journals?

I really appreciate your help with this.

I do not think there is an effective limit on the number of files that can be in a single journal. We journal all JDE related libraries to the same journal.
We have not experienced any performance issues that can be attributed to journalling. You should make sure your hardware is sized appropriately as of course journal will take additional resources. Mimix can have many journal definitions. Each data group would be associated with a different journal.

How do you guys handle the machine name for the two machines?
How do you handle the cut-over, DNS entry or IP address?

We changed the name of our new iSeries and now I have to change the WebSphere cell, node & server. Server I have already renamed but I'm still looking for help with cell & node commands.

Should these be renamed with wsadmin commands?

We have the database named the same on both boxes (WRKRDBDIRE, *LOCAL entry) in the case of a failure, we will use DNS to 'rename' the machine for the other servers. We have done a successfull role swap test. One thing to bear in mind, you want to replicate your JAS servers as well if your not all in one. As JDE is way to chatty to have in a different location than your DB server.

HTH, Tom

Thanks Tom.
I plan to move all our JAS to the iSeries. We currently have 2 instances on iSeries and 7 on Windows.

We are upgrading our 570 with 2 740s and MIMIX. We renamed our new boxes XXXXX1 & XXXXX2 with corresponding RDB entries. Will this cause problems during cut-over from one machine to the other?

It appears our hardware/MIMIX installer is unfamiliar with JDE/E1/MIMIX requirements.

We did not change server names.
We did the PD cutover over the weekend when JDE was not available to the end users.
I've never used the wsadmin commands to rename cell or node names.
We do not use IBM i for Websphere. We use Windows servers.

Cody,

It is the RDB name that matters. So they need the same one. The 'system name' and DNS names don't really matter.

Tom

Sorry to bump an old thread but I had a question on Mimix and JDE Packages. Do you replicate the oneworld/packages folder on the AS400? Replicating these object is a must, correct?

Also, am I right to assume to that to given 2 identical machines when the disaster happens for the backup box to work the relation db name would have to match source machine, perhaps change the host table to point the source hostname to the DR box ip, and route DNS for the source box to point to backup DR box? Given all this is done correctly things should work? Will things like package build/deploys work fine on the DR box if it had to be used long term?

Sorry for double post, I wrote the above late last night and it didn't make sense!

Sorry to bump an old thread but I had a question on Mimix and JDE Packages. Do you replicate the oneworld/packages folder on the AS400? Replicating these objects are a must, correct?

Also, am I right to assume that given 2 identical machines when the disaster happens for the backup box to work the relational db name would have to match source machines relational db name, and you would need to change the host table on backup box to have source hostname pointing to the DR box ip, and then of course route DNS for the source box to point to backup DR box?

Given all this is done correctly things should work? Will things like package build/deploys work fine on the DR box if it had to be used long term? As the 2 iSeries would have 2 different machine names I am not sure if there are any implications? Or as far as JDE is concerned the machine name is whatever DNS resolves to?

[ QUOTE ]
Do you replicate the oneworld/packages folder on the AS400? Replicating these objects are a must, correct?

[/ QUOTE ]

We replicate that directory, yes, but I don't think it is a must (someone can correct me if I'm wrong). You only need it if you plan on building update packages or redeploying an existing package when running on the backup box

[ QUOTE ]
Also, am I right to assume that given 2 identical machines when the disaster happens for the backup box to work the relational db name would have to match source machines relational db name

[/ QUOTE ]

Correct, the *LOCAL entry would be the same name for both boxes which would be the source box.

[ QUOTE ]

, and you would need to change the host table on backup box to have source hostname pointing to the DR box ip, and then of course route DNS for the source box to point to backup DR box?

[/ QUOTE ]

You shouldn't have to change any host table entries on either box. Each box will have two static IP addresses, one for talking on the network in general so JDE and clients can get to them and one dedicated for Mimix replication and configuration between the two. During the switch process you will do one of two things, you will either 1) enable/disable ethernet addresses on each box so that JDE and clients will get routed to the backup box when referencing the normal production server name or 2) change the IP addresses for the source in standard DNS. Your Mimix consultant or professional services person will be able to best determine what is the best option.

[ QUOTE ]
Given all this is done correctly things should work? Will things like package build/deploys work fine on the DR box if it had to be used long term?

[/ QUOTE ]
Don't forget the business data! Building packages when running on the backup server is probably the only thing I haven't tested but we have it setup that we should be able to. All JDE functions work including 3rd party apps connecting directly into JDE when running on the backup box.

[ QUOTE ]

As the 2 iSeries would have 2 different machine names I am not sure if there are any implications? Or as far as JDE is concerned the machine name is whatever DNS resolves to?

[/ QUOTE ]

As I said before make sure both *LOCAL RDB entries are the same on both machines and that they are the same as the source machine name. The server names can stay different even when in a failover state.

Thanks very much for your reply. Certainly clears some things up in my head. Regarding the oneworld/packages folder I can't really see why it would be needed as I believe the files only exist here when JDE compiles them during a build?

One last question. I have read some mimix documents that say to end replication on the oneworld/packages if enabled as it may cause issues with package builds. Is this still something people do, or is it a older problem that's not needed anymore?

It was an issue a couple releases ago with Mimix 5 (and other mirroring products as well), but only with full packages. I haven't tried it since then now that we are running the latest version, 7.1, but I only build about two full production packages a year, the rest are updates, so it's really no extra work just to end the data group befoer kicking off the full package and starting it up afterwards. In the past the package would hang when transfering to the server build.

We have a DG called packages that contains the build libraries & directories.
I have built & deployed update & full packages without stopping the data groups and it went just fine. We are 8.12/8.97.1.1 & Mimix 7.1.16.

[ QUOTE ]
We have a DG called packages that contains the build libraries & directories.
I have built & deployed update & full packages without stopping the data groups and it went just fine. We are 8.12/8.97.1.1 & Mimix 7.1.16.

[/ QUOTE ]

Thanks for the reply. What in total are you replicating for the Packages? When you say the packages libraries do you just mean the "PD812" library and then IFS oneworld/package folder?

Also, are you both using the DNS strategy? Where in the event of a disaster everything is routed correctly via a DNS change as opposed to the backup box being configured with the IP of the production box? Our boxes will be in different states so having the same IP addresses would be a lot of work.

Our two servers are not that far from each other right now so they are currently in the same subnet so we assign the the backup box the IP address of primary during the fail over process and disable that IP on the primary until we fail back, then we just undo what was done. These steps are all automated as well in the step processes. Once we get the two servers on a different subnet we will go the route of a DNS change during failover instead.

Junior,
I'm curious about your JDE configuration with Mimix. We have Mimix installed and replecating for 6 months but can't pass the switch test. When we try to switch to the DR box users can't log into JAS, error "security server is down".

Our JDE setup is follows:
IBM Power 720 Series - Identical OS=V6R1
All-in-One conviguration with 10 JAS on WebSphere 6.1
Boxes are on different subnets (PD=xxx.3.xxx.xxx & DR=xxx.2.xxx.xxx)
Relational Database Name is same on both
Host name is different (PD=XXXX1 & DR=XXXX2) on the same domain
JDE.ini, JAS.ini & JDBJ.ini files are all same on both boxes

Should they be on different ports? ie. PD=6014 & DR=6015

I'm running out of ideas so any help you can give is appreciated.