Miverson

Innovator w/ @Optum, Husband to @mar_iverson, #platforms #tinkererofthings #technology photocred: @jordanimal

[Miverson @Iversonm]

I have an idea @armon @HashiCorp

What if for #Consul and partitions, we allow for different servers to be the leader at the partition level. Essentially that would allow for Consul to be scaled out extremely large without worrying about if a single Consul server can handle all the gossip/raft traffic but still sharing the consul server nodes.

Then if a server went offline, only a portions of the partitions need to re-elect a new leader. 5 partitions and 5 servers? They each lead a partition. 1000 partitions and 20 servers? They each handle 50 partition leadership. It would allow for Consul to scale to millions of clients easily

This idea kept me awake last night

[Armon Dadgar @armon]

@Iversonm @HashiCorp Nomad installer is in the works! Main reason is to reduce additional dependencies.

[Miverson @Iversonm]

Okay I have another question for @armon and @HashiCorp

Why the heck does the Terraform Enterprise install use docker directly instead of running it all through Nomad?? I feel like we should be eating our own dog food #terraform #nomad

[HashiCorp, an IBM Company @HashiCorp]

As @Roblox grows and evolves rapidly, using HashiCorp Nomad, a scheduler and workload orchestrator, enables them to scale their global gaming platform easily and reliably. https://t.co/2mOzWG3D0Z

[HashiCorp, an IBM Company @HashiCorp]

Configure HashiCorp #Consul’s transparent proxy on virtual machines to find and connect to services in the service mesh with DNS. https://t.co/3QTXOGyL4k

[HashiCorp, an IBM Company @HashiCorp]

In this blog, learn how to configure transparent proxy on virtual machines to connect to services in HashiCorp #Consul service mesh. https://t.co/3QTXOGyL4k

[Armon Dadgar @armon]

@Iversonm @HashiCorp For sure! Definitely some fun distributed systems trade offs. Another approach is to run most nodes without fsync or on ramdisk and trust the replicated nature, since Raft guarantees at least N/2 replicas. We’ve seen deployments with that type of config for very high throughputs.

[Miverson @Iversonm]

@justinmchase @alexellisuk @openfaas @HashiCorp A @HashiCorp product that is super BA

https://t.co/AGzpbUUBwN

[Miverson @Iversonm]

@alexellisuk @openfaas
What would take to convince you to start adding @HashiCorp #Nomad support back int openfaas?

[Miverson @Iversonm]

That makes sense around the complexity which is why I was thinking it would only be a thing for large enterprise clusters. I agree on the WAL but now you are depending on disk fsyncs which can have random issues depending on other non controllable environmental factors. Noisy neighbors. Bad hardware. Backups. Or failing disks from an array and seeing a substantial decrease in IOPs as the array repairs itself. Which could bring everything to a screeching halt.

Not arguing, just thinking out loud

[Miverson @Iversonm]

I have an idea @armon @HashiCorp

What if for #Consul and partitions, we allow for different servers to be the leader at the partition level. Essentially that would allow for Consul to be scaled out extremely large without worrying about if a single Consul server can handle all the gossip/raft traffic but still sharing the consul server nodes.

Then if a server went offline, only a portions of the partitions need to re-elect a new leader. 5 partitions and 5 servers? They each lead a partition. 1000 partitions and 20 servers? They each handle 50 partition leadership. It would allow for Consul to scale to millions of clients easily

This idea kept me awake last night

[Miverson @Iversonm]

@SilverEagleDev We are working on moving certain workloads and getting a ton of positive feedback from the business side on how easy, stable and fast it is

[Miverson @Iversonm]

Because I think we are seeing an uptick in people frustrated with the complexities of K8s and Nomad is picking up steam.

I would love to adopt @openfaas at @Optum but don’t want to build out more K8s as it requires substantial labor to manage

[Miverson @Iversonm]

You could then take that logic and add it to the new #Nomad 1.6 release that allows for clients to be bound to a specific namespace

You could also use it for leadership with Vault and root level namespaces. It could literally help most of the @HashiCorp suite of products

[Miverson @Iversonm]

@aantn Yeah. Well ** to nomad. K8s, openshift, etc

We have even successfully built and deployed nomad clients to the mainframe. It was pretty cool seeing a single allocation with 1.5TB of ram and 2.8Thz of cpu allocated

[Miverson @Iversonm]

@Kian_NH @yacinehmito @HashiCorp @Cloudflare Also interested in this. Can like get some visibility with Hashicorp if you don’t have connections already

[Miverson @Iversonm]

I can’t even complain. I’m happy to pay it. Shit is taken care of. Our kids and schools are getting extra funding. And compared to the federal tax I pay, the state tax is peanuts but always feels to make a bigger impact.

Take that surplus and spend it on shit. Then do it again for the next 5 years. Already one of the best states looking at multiple values like graduation rates, employment rates and HDI

[Jeremy Evans @jeremyevans0]

Sequel 5.67.0 released: 20-40% increase in dataset performance on Ruby 2.4+ by avoiding the use of singleton classes, set_literalizer extension for treating sets similar to arrays, schema dumper fixes on MSSQL https://t.co/axXfgtK4BD

[HashiCorp, an IBM Company @HashiCorp]

As @Roblox grows and evolves rapidly, using HashiCorp Nomad, a scheduler and workload orchestrator, enables them to scale their global gaming platform easily and reliably. https://t.co/2mOzWG3D0Z

[Rose W @rose_w]

@Iversonm @ShopifyEng I’m seeing what I can find out for an owner.