For some background on what’s below, see Part 1 of this interview.
Without further ado, let’s continue into part 2!
Integration challenges – there’s a classic lower common denominator challenge. In the past I worked with VCE where putting a layer on top of just a “few things” (servers, storage, compute, hypervisors) was really hard and the results weren’t great.
What’s the fundamental difference in how Terraform handles this vs. other products that have tried and failed?
Andrew Note: we partially covered this previously in the Armon + Mitchell interview where Mitchell stated that “I’m ‘militantly clear’ on this. We don’t abstract cloud providers or hide capabilities.”
Our partner agreement helps with this – part of it is that HashiCorp will run tests but the provider has to write them. These go beyond just unit testing into testing under load – it’s a significant spend by HashiCorp in running these tests. While a provider can trigger a new release, HashiCorp has to approve it.
Going one layer down – let’s look application / API versioning layers. An example is Azure and how different regions had different capabilities or different regions might be on different SDK versions. The Terraform provider was one of the only things out there calling multiple regions and/or and running into this issue.
We worked closely with the Azure SDK team to find a better answer than just documenting and kicking back to the customer. Now the Azure SDK changes what features are allowed based on the region requested.
What’s your favorite new feature? Or, what’s something you’d like to talk about we’ve not covered yet?
I’m excited about my wedding next week!
From a tech standpoint, I’m excited by what the broader service mesh represents for the industry. It will improve application security in a big way. This isn’t cloud specific – will scale down to private datacenters as well.
Overall, cloud has shattered stagnation. In a non-cloud world Vault probably wouldn’t have succeeded – ssh key management wasn’t painful enough. Cloud created much higher demands around making security dynamic.
Inflection Points – looking back, what are some forks in the road where looking back it was a major decision that played out well or not well?
The biggest successful inflection point was hiring Dave McJannet. After making the decision to focus on enterprise, hiring him was critical.
There’s one that went wrong – a product named Atlas. We built a product to unify all our products. Unfortunately, it was expensive and a fairly long mistake we had to support for years. We just ripped out the final Atlas code last month. Simply put, customers didn’t want it – issues with separate products, SKU’s, pricing issues, and more.
Nomad – I’ve heard it’s the best in the industry for container scheduling and managing service mesh but it doesn’t have the marketshare of others. How you do see that improving?
Andrew Note: I had this as a question but it was heavily discussed in the keynote – how Nomad can be complementary to other items in the space. I won’t try to recap that keynote section here – suffice it to say that Mitchell and HashiCorp see this as a solvable issue and have a clear plan.
Scaling & Experience Dilution – coming from a company that grew 6-7x in 2 years (200 to 1400 people), I know there’s no easy answer. How do you handle this personally and as a company?
We have an event every year called Hex – our internal conference. The keynote this year was all 3 of us (Armon, Dave, & Mitchell) talking about this exact topic. The only way to preserve our culture is to stay focused on the Tao. This is a very personal thing for me as well. It’s both a company focus and something that Armon and I monitor and do periodically “swoop in on”.
We’re trying to pay the price upfront to cement this as the company grows. The 200 employees last year will be the core of the company when it gets to 1500.
We also discuss the Tao and our culture during the interview process, annual reviews, etc. The goal is to make it pervasive.
What do you see on a 3 year horizon whether HashiCorp, cloud, or IT in general? (as they all interconnect)
Andrew Note: this was originally my final question for Mitchell. Since I asked it earlier when talking with both Armon & Mitchell (ee “HashiConf 2018 – Interview with Armon & Mitchell”), I didn’t ask it here.
That’s a wrap for part 2.
Thanks for reading!
Disclaimer: I attended HashiConf on a free media pass however paid for my own flights and hotel. There was no requirement for me to blog about any of the content presented and I was not compensated for my time at the event (unless random booth swag from sponsors counts). No materials discussed were presented under NDA.
One thought on “HashiConf 2018 – Interview with Mitchell Hashimoto, Part 2”
Pingback: HashiConf 2018 – Interview with Mitchell Hashimoto, Part 1 | Think Meta