Correct we are already trying it with nossl. We haven't tried targeting just one TS7700, but both are setup pretty much identical. The only difference was that in our second TS7700 the Dasd IP addresses were in a different order, that has been changed, so now both machines are in the same order.
Original Message:
Sent: Wed April 19, 2023 01:25 PM
From: Lourie Goodall
Subject: Implementation of Transparent Cloud Tiering.
Yes, will definitely need support to help as it may be due to a number of things.
It does not matter whether you left your primary DASD configured or not configured to the object store. Each DS8k is independent and not aware of the other's connection. When you hyperswap, it simply starts using the new 'primary' DS8k to route the commands through. Although, it would be best practice to leave them configured so that TCT seamlessly continues to operate after a hyperswap. This is true for your Global Mirror as well.
Depending on how you set up your TS7700 grid, you would normally target your primary and MM to the same TS7700 clusters on your local site and your GM is usually configured to TS7700 at that local site.
Just to ensure I understand your configuration...you are targeting TS7700s and are those TS7700s using FC5283 Advanced Object Store or are you using the older FC5282?
I would follow what your support is asking you to do but a couple of things you could try is if you are running mkcloudserver to two TS7700s, you should try to target just one and see if that works. Second, it looks like you are already doing it but use -nossl in the command for now. That takes out any certificate issues.
------------------------------
Lourie Goodall
Original Message:
Sent: Wed April 19, 2023 09:53 AM
From: John Benik
Subject: Implementation of Transparent Cloud Tiering.
Hi Lourie:
Yes that is what I am referring to The B dasd or metro mirror copy. we are getting this error.
We do have a ticket open, and I have also been trying to open a new discussion on this and not having any luck.
I'll reach out to support again on that. A question did come up the other day and I don't believe we tried it yet. Once successful on the A or primary dasd with the MKCLDSERVER we do a removal and then attempt it on the B dasd. Does it need to stay in place? If so it would stand to reason that when we attempt our C or global mirror copy should also leave A and B in place or at a minimum the A dasd since that's the primary.
------------------------------
John Benik
Original Message:
Sent: Mon April 17, 2023 06:52 PM
From: Lourie Goodall
Subject: Implementation of Transparent Cloud Tiering.
Hello John,
When you say A and B...are you referring to primary vs metro mirror?
If so, you went through the same setup steps as you did on your primary...configuring the Ethernet ports, setting up the routing table and entering the metro mirror DS8K CEC IP Addresses to the TS7700 (SSR activity)?
If you think you have all the steps completed and the network is good between the MM system and TS7700 then you may have to open a ticket to have this analyzed. The number one reason we see issues during the initial set up is typically due to a network issue or configuration step missing.
------------------------------
Lourie Goodall
Original Message:
Sent: Mon April 17, 2023 09:09 AM
From: John Benik
Subject: Implementation of Transparent Cloud Tiering.
Can you tell me if this includes both A and B dasd or just the A dasd? We have attempted the mkcldserver command on our A dasd and had success, however when we try the same command on our B dasd we have gotten errors, but haven't tried after some recent fixes.
------------------------------
John Benik
Original Message:
Sent: Fri April 14, 2023 11:59 AM
From: Lourie Goodall
Subject: Implementation of Transparent Cloud Tiering.
Yes, and if you really want to get fancy you could show the coorelation between the cloud constructs and how the data would be routed to reach respective cluster through that cluster's grid links:
Z sends the command to the DS8k including the cloud construct name (Cloudname1 or Cloudname2) using the same route. DS8k knows the target object store IP Address ranges by the construct name and routes to the respective cluster's grid link IP addresses.
------------------------------
Lourie Goodall
Original Message:
Sent: Fri April 14, 2023 09:57 AM
From: Denis Augusto Pereira
Subject: Implementation of Transparent Cloud Tiering.
Lourie and Team,
As discussed here and a small part in parallel, I did an adjust in the topology, removing Load Balancer between Z and Storage HMCs and adding the second grid, since it will be the final architecture. Please, again, can you take a look?
Topology:
------------------------------
Denis Augusto Pereira
Original Message:
Sent: Thu April 13, 2023 08:09 PM
From: Lourie Goodall
Subject: Implementation of Transparent Cloud Tiering.
Sounds good....and as Glenn stated, sounds like the Redbook is incorrect and a load balancer should not be used between Z and the HMCs. They will need to correct that in the book.
Yes, you can target up to 8 grids from one DS8K by using DS8k multi-cloud...so you would create a different Cloud Network Connection Construct in the ISMF panel and when you configure on the DS8k using mkcloudserver, you run mkcloudserver twice...one for each construct with the IP addresses for the clusters in each grid.
------------------------------
Lourie Goodall
Original Message:
Sent: Thu April 13, 2023 10:36 AM
From: Denis Augusto Pereira
Subject: Implementation of Transparent Cloud Tiering.
Hi Lourie,
You're right and yesterday I create the topology below, where the TCT data will be stored in 2 TS7770 near and the 3rd will only receive the objects by replication of cluster 1 or cluster 2. Please, take a look:
------------------------------
Denis Augusto Pereira
Original Message:
Sent: Wed April 12, 2023 08:18 PM
From: Lourie Goodall
Subject: Implementation of Transparent Cloud Tiering.
Hey Denis,
That looks about right. Only other comment is if they have the ability to include a second TS7700 under the same Cloud Network Connection Construct connection. They can additionally load balance between two TS7700s in the grid but, more importantly, if they lose the one cluster then they have no other entry point into the grid. The second cluster allows them to access if one cluster is down.
------------------------------
Lourie Goodall
Original Message:
Sent: Wed April 12, 2023 01:45 PM
From: Denis Augusto Pereira
Subject: Implementation of Transparent Cloud Tiering.
Glenn, Lourie,
Please, can you evaluate if the topology below is correct?
------------------------------
Denis Augusto Pereira
Original Message:
Sent: Tue April 11, 2023 03:49 PM
From: Glenn Wilcock
Subject: Implementation of Transparent Cloud Tiering.
Hi Denis,
For the load balancer, DFSMS authenticates with an HMC and gets a token that it then uses for requests. Each HMC has a unique token. A load balancer does not know which requests go to which HMC. So, you would get authentication failures because the load balancer would not know to direct a specific request to a specific HMC based on which HMC provided the token. If there is a failure with an HMC, the client has to manually change to another HMC. It's in the backlog to have a more resilient solution.
------------------------------
Glenn Wilcock
Original Message:
Sent: Tue April 11, 2023 12:40 PM
From: Denis Augusto Pereira
Subject: Implementation of Transparent Cloud Tiering.
Hi Glenn,
We're well, thanks, hope you and yours are well too!
As I said to the Lourie, firstly, many thanks of your support. I'm glad that my post here found you and her.
The DS8K links are 10Gb links. I will copy the category of FIXCAT to share with my customer.
About a load balancer between Mainframe and DS8K, I shared above with Lourie 2 situations were customer could face a HMC unavailability, if customer do not have a load balancer, how the solution will work when they face a problem or a scheduled unavailability?
Other doubt to you, although you said the OAM support standard S3 protocols and I believe all DFSMS use case of TCT probably support any standard S3 protocols, how about the support? I mean, if customer implement a solution with, for example, Ceph or Dell ECS, that today are not solutions that was tested, if customer face a problem, does the IBM support can say that the trigger was the not tested solution, do you agree? Or do I am wrong about it?
Many thanks.
------------------------------
Denis Augusto Pereira
Original Message:
Sent: Thu April 06, 2023 10:25 AM
From: Glenn Wilcock
Subject: Implementation of Transparent Cloud Tiering.
Hi Denis, hope that all is well!
For the DS8K ethernet links, we recommend that they upgrade to the 10Gb links, if they have not done so.
For the DFSMS support, they should apply all current maintenance for TCT, using the following FIXCAT category: IBM.Function.DFSMSCloudStorage and DFSMSCS/K
For TCT, they need ethernet between Mainframe <> DS8K <> TS7700 and/or Object Storage.
For OAM cloud support, they need ethernet from Mainframe to Object Storage. (OAM cannot write to the object store of the TS7700).
The Redbook is downlevel. A Load Balancer for the ethernet shouldn't be utilized.
TCT / OAM support S3 protocols. Object Storage providers that support standard S3 protocols are supported.
------------------------------
Glenn Wilcock
Original Message:
Sent: Tue April 04, 2023 02:29 PM
From: Denis Augusto Pereira
Subject: Implementation of Transparent Cloud Tiering.
Hello Experts,
Please, I would like to ask help / support to validate some information related to the implementation of Transparent Cloud Tiering.
To the hardware side, customer already have the ethernet ports in the DS8K (FC 3603) and they also have TS7700 Advanced Object Store (FC 5283, 8083 and FC 3479).
To the software side, I attached a PDF where I list the corrections/PTFs necessaries to implement the TCT, based in a Redbook (SG24-8381).
To the network / connectivity side, I understood that are necessary ethernet connections between:
- Mainframe and DS8K
- DS8K and TS7700
- Mainframe and TS7700 (use case of OAM)
According to the Redbook (SG24-8381) and implementations of object storage on-prem (COS) that my customers bought, looks like a requirement the utilization of a Load Balancer, for performance and availability.
Please, do I need something additional of what was described here?
Other doubt, although few object storage solutions were homologated to be an object repository of TCT (IBM COS on-prem, IBM Cloud - COS, TS7700, RStor and AWS), customer would like to know if other solutions can be used, like Hitachi HCP, Dell ECS, Cloudian and Ceph. Please, does these solutions can be used as an object storage target of our TCT implementation?
Many thanks in advance.
------------------------------
Denis Augusto Pereira
------------------------------