redis-cluster | redis ha cluster -

Is that true or Am I missing someting?

The error message is correct. You can't replace named resources. Recent AWS blog post explains how to deal with such a situation:

• How do I resolve the AWS CloudFormation error "Cannot update a stack when a custom-named resource requires replacing"?

And the resolution is that you have to rename your resource sadly, or remove name, so you will have to create new cluster.

But either way, since your update requires replacement you will always get new cluster. Its only the question whether it will have different name or same. So you have to backup it first, and then restore to newly created one.

I realized I needed to open port 8443 from the documentation.

So I went to the firewall list in google cloud. Found the rules that had tcp:80,443 in the Protocols / ports. Clicked it, clicked edit and added 8443 to it.

I had an error after but this fixed it:

Generally it is possible to do it through import procedure as explained in Moving resources between stacks.

However, AWS::ElastiCache::CacheCluster does not support import procedure. Only some resources support it. Sadly, ElastiCache is not one of these resources.

So if you don't want to touch your prod cluster, for now you are stuck with what you have. Maybe in a near future support for importing, and subsequently moving AWS::ElastiCache::CacheCluster will be added to CloudFormation. Otherwise you would have to snapshot it, and recreated in a new stack.

I was missing spec.endpoints.path on the ServiceMonitor

Here is an example manifest from adding new scraping targets and troubleshooting tutorial.

There're several solution you can try:

• You can use the rename-command config to disable write commands. If you only want to disable small number of commands, that's a good solution. However, since there're too many write commands, you might need to have too many configuration, and easy to miss some of them.

• If you're using Redis 6.0, you can use Redis ACL to disable write commands for specific users.

• You can setup a read-only Redis replica for your master, and ask clients to read from the replica.

The example is of the old client code,. After hours of toil I found that the following works:

From an Azure perspective, you have two options and if you are very specific to option two but are looking for the Sentinel architecture piece, there is business continuity and high availability options in both IaaS (Linux VM scale sets) and PaaS services that go beyond the Sentinel component.

1. Azure Cache for Redis (PaaS) where you choose & deploy your desired service tier (Premium Tier required for HA) and connect your client applications. Please see: Azure Cache for Redis FAQ and Caching Best Practice.
2. The second option is to deploy a solution (as you have detailed) as an IaaS solution built from Azure VMs. There are a number of Redis Linux VM images to choose from the Azure Marketplace or there is the option to create a Linux VM OS image from your on-premise solution and migrate that to Azure. The Sentinel component is enabled on each server (master, slavea, and slaveb, ...). There are networking and other considerations too. For building a system from scratch, please see: How to Setup Redis Replication (with Cluster-Mode Disabled) in CentOS 8 – Part 1 and How to Setup Redis For High Availability with Sentinel in CentOS 8 – Part 2

Based on the logs

- ingress.extensions/ingress-fibonacci-service configured

It sounds like your application is served by an ingress so it's likely that it's going through a path that is not /. You can see what you have in the ingress.

You can check the configs of the ingress and logs of the ingress controller (depending on what you are using)

Unfortunately at this time there is no AWS Solution for master-master within ElastiCache.

Here are some of the solutions you could do:

• Within your application use write endpoint and read endpoint for Redis, the write endpoint would target the primary write global datastore, the read would target the local read datastore
• Write to a local region resource (such as DynamoDB), have Lambda trigger and write to the primary write global datastore.
• Write to a local region queue/stream such as SQS or Kinesis Data Streams (which supports a Lambda consumer) and have Lambda consume and write.

Any of the solutions will have a degree of latency between, option 1 does a direct write whereas the other 2 allow the process to happen in the background. The decision here would be based on your application (can it wait for the write to be done, or dos it need to happen in the flow).

Additionally if this is for writing a cache, could you just use cache priming instead? By this I mean generating a flattened cache via a script to ensure your redis cluster is always up to date.